diff --git a/.gitlab/e2e.yml b/.gitlab/e2e.yml
index 55867ca6859..dadd1befe23 100644
--- a/.gitlab/e2e.yml
+++ b/.gitlab/e2e.yml
@@ -18,14 +18,17 @@
     GROUND_TRUTH_ALPINE_IMAGE: registry.ddbuild.io/alpine:latest
     GROUND_TRUTH_MILLSTONE__IMAGE: ${SALUKI_IMAGE_REPO_BASE}/millstone:${CI_COMMIT_SHA}
     GROUND_TRUTH_DATADOG_INTAKE__IMAGE: ${SALUKI_IMAGE_REPO_BASE}/datadog-intake:${CI_COMMIT_SHA}
-    GROUND_TRUTH_COMPARISON__IMAGE: ${SALUKI_IMAGE_REPO_BASE}/bundled-agent-adp:${CI_COMMIT_SHA}
   script:
     - make build-ground-truth
     - target/release/ground-truth ./test/correctness/${CORRECTNESS_TEST_CASE}/config.yaml
 
-# Mixin for correctness tests where baseline is the CI-built bundled ADP image.
-# Tests with a different baseline (e.g. DDOT) should NOT extend this and instead
-# let config.yaml specify the baseline image directly.
+# Mixin that overrides the comparison image to the CI-built ADP image.
+# Tests comparing against external agent-dev images should NOT extend this.
+.test-correctness-adp-comparison:
+  variables:
+    GROUND_TRUTH_COMPARISON__IMAGE: ${SALUKI_IMAGE_REPO_BASE}/bundled-agent-adp:${CI_COMMIT_SHA}
+
+# Mixin that overrides the baseline image to the CI-built ADP image.
 .test-correctness-adp-baseline:
   variables:
     GROUND_TRUTH_BASELINE__IMAGE: ${SALUKI_IMAGE_REPO_BASE}/bundled-agent-adp:${CI_COMMIT_SHA}
@@ -94,35 +97,72 @@ build-bundled-agent-adp-image:
       .
 
 test-correctness-dsd-plain:
-  extends: [.build-common-variables, .test-correctness-definition, .test-correctness-adp-baseline]
+  extends: [.build-common-variables, .test-correctness-definition, .test-correctness-adp-comparison, .test-correctness-adp-baseline]
   variables:
     CORRECTNESS_TEST_CASE: dsd-plain
 
 test-correctness-dsd-origin-detection:
-  extends: [.build-common-variables, .test-correctness-definition, .test-correctness-adp-baseline]
+  extends: [.build-common-variables, .test-correctness-definition, .test-correctness-adp-comparison, .test-correctness-adp-baseline]
   variables:
     CORRECTNESS_TEST_CASE: dsd-origin-detection
 
 test-correctness-otlp-metrics:
-  extends: [.build-common-variables, .test-correctness-definition, .test-correctness-adp-baseline]
+  extends: [.build-common-variables, .test-correctness-definition, .test-correctness-adp-comparison, .test-correctness-adp-baseline]
   variables:
     CORRECTNESS_TEST_CASE: otlp-metrics
 
 test-correctness-otlp-traces:
-  extends: [.build-common-variables, .test-correctness-definition, .test-correctness-adp-baseline]
+  extends: [.build-common-variables, .test-correctness-definition, .test-correctness-adp-comparison, .test-correctness-adp-baseline]
   variables:
     CORRECTNESS_TEST_CASE: otlp-traces
 
 test-correctness-otlp-traces-ottl-filtering:
-  extends: [.build-common-variables, .test-correctness-definition]
+  extends: [.build-common-variables, .test-correctness-definition, .test-correctness-adp-comparison]
   variables:
     CORRECTNESS_TEST_CASE: otlp-traces-ottl-filtering
 
 test-correctness-otlp-traces-ottl-transform:
-  extends: [.build-common-variables, .test-correctness-definition]
+  extends: [.build-common-variables, .test-correctness-definition, .test-correctness-adp-comparison]
   variables:
     CORRECTNESS_TEST_CASE: otlp-traces-ottl-transform
 
+test-correctness-dsd-tag-filterlist:
+  extends: [.build-common-variables, .test-correctness-definition, .test-correctness-adp-comparison]
+  variables:
+    CORRECTNESS_TEST_CASE: dsd-tag-filterlist
+
+test-correctness-dsd-optimized-tag-filterlist:
+  extends: [.build-common-variables, .test-correctness-definition]
+  needs:
+    - build-datadog-intake-image
+    - build-millstone-image
+  variables:
+    CORRECTNESS_TEST_CASE: dsd-optimized-tag-filterlist
+
+test-correctness-dsd-additional-optimized-tag-filterlist:
+  extends: [.build-common-variables, .test-correctness-definition]
+  needs:
+    - build-datadog-intake-image
+    - build-millstone-image
+  variables:
+    CORRECTNESS_TEST_CASE: dsd-additional-optimized-tag-filterlist
+
+test-correctness-dsd-tag-filterlist-context-cache:
+  extends: [.build-common-variables, .test-correctness-definition]
+  needs:
+    - build-datadog-intake-image
+    - build-millstone-image
+  variables:
+    CORRECTNESS_TEST_CASE: dsd-tag-filterlist-context-cache
+
+test-correctness-dsd-tag-filterlist-post-aggr:
+  extends: [.build-common-variables, .test-correctness-definition]
+  needs:
+    - build-datadog-intake-image
+    - build-millstone-image
+  variables:
+    CORRECTNESS_TEST_CASE: dsd-tag-filterlist-post-aggr
+
 test-integration:
   stage: e2e
   tags: ["docker-in-docker:amd64"]
diff --git a/Cargo.lock b/Cargo.lock
index 78aab11c87a..ee800523c8b 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -1100,7 +1100,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "39cab71617ae0d63f51a36d69f866391735b51691dbda63cf6f96d042b63efeb"
 dependencies = [
  "libc",
- "windows-sys 0.52.0",
+ "windows-sys 0.59.0",
 ]
 
 [[package]]
@@ -3035,7 +3035,7 @@ dependencies = [
  "once_cell",
  "socket2 0.6.2",
  "tracing",
- "windows-sys 0.52.0",
+ "windows-sys 0.59.0",
 ]
 
 [[package]]
@@ -3443,9 +3443,9 @@ checksum = "f87165f0995f63a9fbeea62b64d10b4d9d8e78ec6d7d51fb2125fda7bb36788f"
 
 [[package]]
 name = "rustls-webpki"
-version = "0.103.9"
+version = "0.103.10"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d7df23109aa6c1567d1c575b9952556388da57401e4ace1d15f79eedad0d8f53"
+checksum = "df33b2b81ac578cabaf06b89b0631153a3f416b0a886e8a7a1707fb51abbd1ef"
 dependencies = [
  "aws-lc-rs",
  "ring",
@@ -3900,7 +3900,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "5b55fb86dfd3a2f5f76ea78310a88f96c4ea21a3031f8d212443d56123fd0521"
 dependencies = [
  "libc",
- "windows-sys 0.52.0",
+ "windows-sys 0.59.0",
 ]
 
 [[package]]
diff --git a/Makefile b/Makefile
index 9fae09d7e2b..6b145b871db 100644
--- a/Makefile
+++ b/Makefile
@@ -521,7 +521,7 @@ test-all: test test-property test-docs test-miri test-loom
 
 .PHONY: test-correctness
 test-correctness: ## Runs the complete correctness suite
-test-correctness: test-correctness-dsd-plain test-correctness-dsd-origin-detection test-correctness-otlp-metrics test-correctness-otlp-traces test-correctness-otlp-traces-ottl-filtering test-correctness-otlp-traces-ottl-transform
+test-correctness: test-correctness-dsd-plain test-correctness-dsd-origin-detection test-correctness-dsd-tag-filterlist test-correctness-dsd-optimized-tag-filterlist test-correctness-dsd-additional-optimized-tag-filterlist test-correctness-dsd-tag-filterlist-context-cache test-correctness-dsd-tag-filterlist-post-aggr test-correctness-otlp-metrics test-correctness-otlp-traces test-correctness-otlp-traces-ottl-filtering test-correctness-otlp-traces-ottl-transform
 
 .PHONY: test-correctness-dsd-plain
 test-correctness-dsd-plain: build-ground-truth
@@ -535,6 +535,36 @@ test-correctness-dsd-origin-detection: ## Runs the 'dsd-origin-detection' correc
 	@echo "[*] Running 'dsd-origin-detection' correctness test case..."
 	@target/release/ground-truth $(shell pwd)/test/correctness/dsd-origin-detection/config.yaml
 
+.PHONY: test-correctness-dsd-tag-filterlist
+test-correctness-dsd-tag-filterlist: build-ground-truth
+test-correctness-dsd-tag-filterlist: ## Runs the 'dsd-tag-filterlist' correctness test case
+	@echo "[*] Running 'dsd-tag-filterlist' correctness test case..."
+	@target/release/ground-truth $(shell pwd)/test/correctness/dsd-tag-filterlist/config.yaml
+
+.PHONY: test-correctness-dsd-optimized-tag-filterlist
+test-correctness-dsd-optimized-tag-filterlist: build-ground-truth
+test-correctness-dsd-optimized-tag-filterlist: ## Runs the 'dsd-optimized-tag-filterlist' correctness test case
+	@echo "[*] Running 'dsd-optimized-tag-filterlist' correctness test case..."
+	@target/release/ground-truth $(shell pwd)/test/correctness/dsd-optimized-tag-filterlist/config.yaml
+
+.PHONY: test-correctness-dsd-additional-optimized-tag-filterlist
+test-correctness-dsd-additional-optimized-tag-filterlist: build-ground-truth
+test-correctness-dsd-additional-optimized-tag-filterlist: ## Runs the 'dsd-additional-optimized-tag-filterlist' correctness test case
+	@echo "[*] Running 'dsd-additional-optimized-tag-filterlist' correctness test case..."
+	@target/release/ground-truth $(shell pwd)/test/correctness/dsd-additional-optimized-tag-filterlist/config.yaml
+
+.PHONY: test-correctness-dsd-tag-filterlist-context-cache
+test-correctness-dsd-tag-filterlist-context-cache: build-ground-truth
+test-correctness-dsd-tag-filterlist-context-cache: ## Runs the 'dsd-tag-filterlist-context-cache' correctness test case
+	@echo "[*] Running 'dsd-tag-filterlist-context-cache' correctness test case..."
+	@target/release/ground-truth $(shell pwd)/test/correctness/dsd-tag-filterlist-context-cache/config.yaml
+
+.PHONY: test-correctness-dsd-tag-filterlist-post-aggr
+test-correctness-dsd-tag-filterlist-post-aggr: build-ground-truth
+test-correctness-dsd-tag-filterlist-post-aggr: ## Runs the 'dsd-tag-filterlist-post-aggr' correctness test case
+	@echo "[*] Running 'dsd-tag-filterlist-post-aggr' correctness test case..."
+	@target/release/ground-truth $(shell pwd)/test/correctness/dsd-tag-filterlist-post-aggr/config.yaml
+
 .PHONY: test-correctness-otlp-metrics
 test-correctness-otlp-metrics: build-ground-truth
 test-correctness-otlp-metrics: ## Runs the 'otlp-metrics' correctness test case
diff --git a/bin/correctness/ground-truth/src/analysis/metrics/types.rs b/bin/correctness/ground-truth/src/analysis/metrics/types.rs
index f0952d6d1a3..a725c0cd563 100644
--- a/bin/correctness/ground-truth/src/analysis/metrics/types.rs
+++ b/bin/correctness/ground-truth/src/analysis/metrics/types.rs
@@ -107,8 +107,8 @@ impl NormalizedMetric {
     /// # Errors
     ///
     /// If the raw values are empty, or if the values cannot be normalized, an error is returned.
-    pub fn try_from_values(
-        context: MetricContext, mut raw_values: Vec<(u64, MetricValue)>,
+    fn try_from_normalized_values(
+        context: NormalizedMetricContext, mut raw_values: Vec<(u64, MetricValue)>,
     ) -> Result<Self, GenericError> {
         // We need to first sort the raw values by timestamp, to ensure we have proper ordering semantics.
         raw_values.sort_by(|a, b| a.0.cmp(&b.0));
@@ -116,7 +116,7 @@ impl NormalizedMetric {
             .with_error_context(|| format!("Failed to normalize values for metric '{}'", context.name()))?;
 
         Ok(Self {
-            context: NormalizedMetricContext::from_stele_context(context),
+            context,
             raw_values,
             value,
         })
@@ -176,23 +176,24 @@ impl NormalizedMetrics {
             return Err(generic_error!("Cannot normalize an empty set of metrics."));
         }
 
-        // Aggregate metric values by (context, type), using a `BTreeMap` so that we can get sorted metrics for ~free.
+        // Aggregate metric values by normalized (context, type), using a `BTreeMap` so that metric ordering is stable
+        // even when the original intake payload emits the same tag set in a different order.
         //
         // We group by type because metrics with the same context but different types (e.g., Count vs Rate) cannot be
         // merged together during normalization.
         let mut aggregated_context_values = BTreeMap::new();
 
         for metric in metrics {
-            let context = metric.context();
+            let context = NormalizedMetricContext::from_stele_context(metric.context().clone());
             let metric_type = metric_value_type(metric.values());
-            let key = (context.clone(), metric_type);
+            let key = (context, metric_type);
             let context_values = aggregated_context_values.entry(key).or_insert_with(Vec::new);
             context_values.extend_from_slice(metric.values());
         }
 
         let metrics = aggregated_context_values
             .into_iter()
-            .map(|((context, _type), values)| NormalizedMetric::try_from_values(context, values))
+            .map(|((context, _type), values)| NormalizedMetric::try_from_normalized_values(context, values))
             .try_fold(Vec::new(), |mut metrics, maybe_metric| {
                 metrics.push(maybe_metric?);
                 Ok::<_, GenericError>(metrics)
@@ -316,3 +317,42 @@ fn try_normalize_values(raw_values: &[(u64, MetricValue)]) -> Result<MetricValue
 
     Ok(current_value)
 }
+
+#[cfg(test)]
+mod tests {
+    use serde_json::json;
+    use stele::Metric;
+
+    use super::NormalizedMetrics;
+
+    fn metric(name: &str, tags: &[&str], points: &[(u64, f64)]) -> Metric {
+        serde_json::from_value(json!({
+            "context": {
+                "name": name,
+                "tags": tags,
+            },
+            "values": points
+                .iter()
+                .map(|(ts, value)| json!([ts, {"mtype": "Count", "value": value}]))
+                .collect::<Vec<_>>(),
+        }))
+        .expect("metric should deserialize")
+    }
+
+    #[test]
+    fn normalizes_context_order_before_grouping_metrics() {
+        let metrics = vec![
+            metric("tagfilter.miss", &["pod:pod-a", "env:prod"], &[(10, 2.0)]),
+            metric("tagfilter.miss", &["env:prod", "pod:pod-a"], &[(20, 3.0)]),
+        ];
+
+        let normalized = NormalizedMetrics::try_from_stele_metrics(&metrics).expect("metrics should normalize");
+
+        assert_eq!(normalized.len(), 1);
+        assert_eq!(
+            normalized.metrics()[0].context().to_string(),
+            "tagfilter.miss[env:prod, pod:pod-a]"
+        );
+        assert_eq!(normalized.metrics()[0].raw_values().len(), 2);
+    }
+}
diff --git a/lib/saluki-components/src/transforms/tag_filterlist/mod.rs b/lib/saluki-components/src/transforms/tag_filterlist/mod.rs
index 4dbadcd22a3..36df58a282d 100644
--- a/lib/saluki-components/src/transforms/tag_filterlist/mod.rs
+++ b/lib/saluki-components/src/transforms/tag_filterlist/mod.rs
@@ -17,7 +17,7 @@ use saluki_core::{
         transforms::{Transform, TransformBuilder, TransformContext},
         ComponentContext,
     },
-    data_model::event::EventType,
+    data_model::event::{metric::Metric, EventType},
     topology::OutputDefinition,
 };
 use saluki_error::GenericError;
@@ -215,7 +215,7 @@ fn apply_tag_filter(tags: &SharedTagSet, is_exclude: bool, names: &HashSet<Strin
 /// If the metric name is not present in `filters`, the metric is left unchanged.
 /// If filtering would not change any tags, the metric context is left untouched (zero allocations).
 #[inline]
-pub fn filter_metric_tags(metric: &mut saluki_core::data_model::event::metric::Metric, filters: &CompiledFilters) {
+pub fn filter_metric_tags(metric: &mut Metric, filters: &CompiledFilters) {
     let Some((is_exclude, tag_names)) = filters.get(metric.context().name().as_ref()) else {
         return;
     };
diff --git a/test/correctness/dsd-additional-optimized-tag-filterlist/config.yaml b/test/correctness/dsd-additional-optimized-tag-filterlist/config.yaml
new file mode 100644
index 00000000000..2ba0d11a92a
--- /dev/null
+++ b/test/correctness/dsd-additional-optimized-tag-filterlist/config.yaml
@@ -0,0 +1,19 @@
+analysis_mode: metrics
+millstone:
+  image: saluki-images/millstone:latest
+  config_path: millstone.yaml
+datadog_intake:
+  image: saluki-images/datadog-intake:latest
+  config_path: ../datadog-intake.yaml
+baseline:
+  image: registry.datadoghq.com/agent:7.76.2-jmx
+  files:
+    - datadog.yaml:/etc/datadog-agent/datadog.yaml
+  additional_env_vars:
+    - DD_API_KEY=correctness-test
+comparison:
+  image: registry.datadoghq.com/agent-dev:62567276-py3-jmx
+  files:
+    - datadog.yaml:/etc/datadog-agent/datadog.yaml
+  additional_env_vars:
+    - DD_API_KEY=correctness-test
diff --git a/test/correctness/dsd-additional-optimized-tag-filterlist/datadog.yaml b/test/correctness/dsd-additional-optimized-tag-filterlist/datadog.yaml
new file mode 100644
index 00000000000..3919762b6c8
--- /dev/null
+++ b/test/correctness/dsd-additional-optimized-tag-filterlist/datadog.yaml
@@ -0,0 +1,48 @@
+# Using a fixed hostname ensures consistent tags between the baseline and comparison agents.
+hostname: "correctness-testing"
+
+# Dummy API key.
+api_key: dummy-api-key-correctness-testing
+
+# We have to specifically configure the health port to use.
+health_port: 5555
+
+# Point ourselves at the datadog-intake service.
+dd_url: "http://datadog-intake:2049"
+
+# Turn off UDP and listen on a UDS socket instead.
+dogstatsd_port: 0
+dogstatsd_socket: /airlock/metrics.sock
+
+dogstatsd_origin_detection: false
+
+# Keep worker count at 1 to ensure deterministic gauge processing.
+dogstatsd_workers_count: 1
+
+metric_tag_filterlist:
+  - metric_name: tagfilter.exclude
+    action: exclude
+    tags:
+      - pod
+  - metric_name: tagfilter.include
+    action: include
+    tags:
+      - env
+  - metric_name: tagfilter.exclude.multi
+    action: exclude
+    tags:
+      - pod
+      - region
+  - metric_name: tagfilter.include.multi
+    action: include
+    tags:
+      - env
+      - service
+  - metric_name: tagfilter.conflict
+    action: include
+    tags:
+      - env
+  - metric_name: tagfilter.conflict
+    action: exclude
+    tags:
+      - pod
diff --git a/test/correctness/dsd-additional-optimized-tag-filterlist/millstone.yaml b/test/correctness/dsd-additional-optimized-tag-filterlist/millstone.yaml
new file mode 100644
index 00000000000..835dbc78aca
--- /dev/null
+++ b/test/correctness/dsd-additional-optimized-tag-filterlist/millstone.yaml
@@ -0,0 +1,74 @@
+seed: [2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61, 67, 71, 73, 79, 83, 89, 97, 101, 103, 107, 109, 113, 127, 131]
+target: "unixgram:///airlock/metrics.sock"
+aggregation_bucket_width_secs: 10
+volume: 20000
+corpus:
+  size: 20000
+  payload:
+    dogstatsd:
+      contexts:
+        constant: 1024
+      metric_names:
+        - "tagfilter.exclude"
+        - "tagfilter.include"
+        - "tagfilter.miss"
+        - "tagfilter.exclude.multi"
+        - "tagfilter.include.multi"
+        - "tagfilter.conflict"
+      tag_names:
+        - "env"
+        - "pod"
+        - "service"
+        - "region"
+        - "version"
+      tag_values:
+        - "prod"
+        - "staging"
+        - "dev"
+        - "pod-a"
+        - "pod-b"
+        - "pod-c"
+        - "web"
+        - "api"
+        - "worker"
+        - "us-east-1"
+        - "us-west-2"
+        - "eu-west-1"
+        - "v1"
+        - "v2"
+        - "v3"
+      name_length:
+        inclusive:
+          min: 1
+          max: 32
+      tag_length:
+        inclusive:
+          min: 3
+          max: 16
+      tags_per_msg:
+        inclusive:
+          min: 2
+          max: 5
+      unique_tag_ratio: 1.0
+      value:
+        float_probability: 0.5
+        range:
+          inclusive:
+            min: -9999999
+            max: 9999999
+      multivalue_count:
+        inclusive:
+          min: 2
+          max: 32
+      multivalue_pack_probability: 0.08
+      kind_weights:
+        metric: 100
+        event: 0
+        service_check: 0
+      metric_weights:
+        count: 0
+        gauge: 0
+        timer: 0
+        distribution: 1
+        set: 0
+        histogram: 0
diff --git a/test/correctness/dsd-optimized-tag-filterlist/config.yaml b/test/correctness/dsd-optimized-tag-filterlist/config.yaml
new file mode 100644
index 00000000000..9205aeeb513
--- /dev/null
+++ b/test/correctness/dsd-optimized-tag-filterlist/config.yaml
@@ -0,0 +1,19 @@
+analysis_mode: metrics
+millstone:
+  image: saluki-images/millstone:latest
+  config_path: millstone.yaml
+datadog_intake:
+  image: saluki-images/datadog-intake:latest
+  config_path: ../datadog-intake.yaml
+baseline:
+  image: registry.datadoghq.com/agent:7.76.2-jmx
+  files:
+    - datadog.yaml:/etc/datadog-agent/datadog.yaml
+  additional_env_vars:
+    - DD_API_KEY=correctness-test
+comparison:
+  image: registry.datadoghq.com/agent-dev:ff848d05-py3-jmx
+  files:
+    - datadog.yaml:/etc/datadog-agent/datadog.yaml
+  additional_env_vars:
+    - DD_API_KEY=correctness-test
diff --git a/test/correctness/dsd-optimized-tag-filterlist/datadog.yaml b/test/correctness/dsd-optimized-tag-filterlist/datadog.yaml
new file mode 100644
index 00000000000..3919762b6c8
--- /dev/null
+++ b/test/correctness/dsd-optimized-tag-filterlist/datadog.yaml
@@ -0,0 +1,48 @@
+# Using a fixed hostname ensures consistent tags between the baseline and comparison agents.
+hostname: "correctness-testing"
+
+# Dummy API key.
+api_key: dummy-api-key-correctness-testing
+
+# We have to specifically configure the health port to use.
+health_port: 5555
+
+# Point ourselves at the datadog-intake service.
+dd_url: "http://datadog-intake:2049"
+
+# Turn off UDP and listen on a UDS socket instead.
+dogstatsd_port: 0
+dogstatsd_socket: /airlock/metrics.sock
+
+dogstatsd_origin_detection: false
+
+# Keep worker count at 1 to ensure deterministic gauge processing.
+dogstatsd_workers_count: 1
+
+metric_tag_filterlist:
+  - metric_name: tagfilter.exclude
+    action: exclude
+    tags:
+      - pod
+  - metric_name: tagfilter.include
+    action: include
+    tags:
+      - env
+  - metric_name: tagfilter.exclude.multi
+    action: exclude
+    tags:
+      - pod
+      - region
+  - metric_name: tagfilter.include.multi
+    action: include
+    tags:
+      - env
+      - service
+  - metric_name: tagfilter.conflict
+    action: include
+    tags:
+      - env
+  - metric_name: tagfilter.conflict
+    action: exclude
+    tags:
+      - pod
diff --git a/test/correctness/dsd-optimized-tag-filterlist/millstone.yaml b/test/correctness/dsd-optimized-tag-filterlist/millstone.yaml
new file mode 100644
index 00000000000..835dbc78aca
--- /dev/null
+++ b/test/correctness/dsd-optimized-tag-filterlist/millstone.yaml
@@ -0,0 +1,74 @@
+seed: [2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61, 67, 71, 73, 79, 83, 89, 97, 101, 103, 107, 109, 113, 127, 131]
+target: "unixgram:///airlock/metrics.sock"
+aggregation_bucket_width_secs: 10
+volume: 20000
+corpus:
+  size: 20000
+  payload:
+    dogstatsd:
+      contexts:
+        constant: 1024
+      metric_names:
+        - "tagfilter.exclude"
+        - "tagfilter.include"
+        - "tagfilter.miss"
+        - "tagfilter.exclude.multi"
+        - "tagfilter.include.multi"
+        - "tagfilter.conflict"
+      tag_names:
+        - "env"
+        - "pod"
+        - "service"
+        - "region"
+        - "version"
+      tag_values:
+        - "prod"
+        - "staging"
+        - "dev"
+        - "pod-a"
+        - "pod-b"
+        - "pod-c"
+        - "web"
+        - "api"
+        - "worker"
+        - "us-east-1"
+        - "us-west-2"
+        - "eu-west-1"
+        - "v1"
+        - "v2"
+        - "v3"
+      name_length:
+        inclusive:
+          min: 1
+          max: 32
+      tag_length:
+        inclusive:
+          min: 3
+          max: 16
+      tags_per_msg:
+        inclusive:
+          min: 2
+          max: 5
+      unique_tag_ratio: 1.0
+      value:
+        float_probability: 0.5
+        range:
+          inclusive:
+            min: -9999999
+            max: 9999999
+      multivalue_count:
+        inclusive:
+          min: 2
+          max: 32
+      multivalue_pack_probability: 0.08
+      kind_weights:
+        metric: 100
+        event: 0
+        service_check: 0
+      metric_weights:
+        count: 0
+        gauge: 0
+        timer: 0
+        distribution: 1
+        set: 0
+        histogram: 0
diff --git a/test/correctness/dsd-tag-filterlist-context-cache/config.yaml b/test/correctness/dsd-tag-filterlist-context-cache/config.yaml
new file mode 100644
index 00000000000..a55bf1d3c74
--- /dev/null
+++ b/test/correctness/dsd-tag-filterlist-context-cache/config.yaml
@@ -0,0 +1,19 @@
+analysis_mode: metrics
+millstone:
+  image: saluki-images/millstone:latest
+  config_path: millstone.yaml
+datadog_intake:
+  image: saluki-images/datadog-intake:latest
+  config_path: ../datadog-intake.yaml
+baseline:
+  image: registry.datadoghq.com/agent:7.76.2-jmx
+  files:
+    - datadog.yaml:/etc/datadog-agent/datadog.yaml
+  additional_env_vars:
+    - DD_API_KEY=correctness-test
+comparison:
+  image: registry.datadoghq.com/agent-dev:aef811fd-py3-jmx
+  files:
+    - datadog.yaml:/etc/datadog-agent/datadog.yaml
+  additional_env_vars:
+    - DD_API_KEY=correctness-test
diff --git a/test/correctness/dsd-tag-filterlist-context-cache/datadog.yaml b/test/correctness/dsd-tag-filterlist-context-cache/datadog.yaml
new file mode 100644
index 00000000000..3919762b6c8
--- /dev/null
+++ b/test/correctness/dsd-tag-filterlist-context-cache/datadog.yaml
@@ -0,0 +1,48 @@
+# Using a fixed hostname ensures consistent tags between the baseline and comparison agents.
+hostname: "correctness-testing"
+
+# Dummy API key.
+api_key: dummy-api-key-correctness-testing
+
+# We have to specifically configure the health port to use.
+health_port: 5555
+
+# Point ourselves at the datadog-intake service.
+dd_url: "http://datadog-intake:2049"
+
+# Turn off UDP and listen on a UDS socket instead.
+dogstatsd_port: 0
+dogstatsd_socket: /airlock/metrics.sock
+
+dogstatsd_origin_detection: false
+
+# Keep worker count at 1 to ensure deterministic gauge processing.
+dogstatsd_workers_count: 1
+
+metric_tag_filterlist:
+  - metric_name: tagfilter.exclude
+    action: exclude
+    tags:
+      - pod
+  - metric_name: tagfilter.include
+    action: include
+    tags:
+      - env
+  - metric_name: tagfilter.exclude.multi
+    action: exclude
+    tags:
+      - pod
+      - region
+  - metric_name: tagfilter.include.multi
+    action: include
+    tags:
+      - env
+      - service
+  - metric_name: tagfilter.conflict
+    action: include
+    tags:
+      - env
+  - metric_name: tagfilter.conflict
+    action: exclude
+    tags:
+      - pod
diff --git a/test/correctness/dsd-tag-filterlist-context-cache/millstone.yaml b/test/correctness/dsd-tag-filterlist-context-cache/millstone.yaml
new file mode 100644
index 00000000000..835dbc78aca
--- /dev/null
+++ b/test/correctness/dsd-tag-filterlist-context-cache/millstone.yaml
@@ -0,0 +1,74 @@
+seed: [2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61, 67, 71, 73, 79, 83, 89, 97, 101, 103, 107, 109, 113, 127, 131]
+target: "unixgram:///airlock/metrics.sock"
+aggregation_bucket_width_secs: 10
+volume: 20000
+corpus:
+  size: 20000
+  payload:
+    dogstatsd:
+      contexts:
+        constant: 1024
+      metric_names:
+        - "tagfilter.exclude"
+        - "tagfilter.include"
+        - "tagfilter.miss"
+        - "tagfilter.exclude.multi"
+        - "tagfilter.include.multi"
+        - "tagfilter.conflict"
+      tag_names:
+        - "env"
+        - "pod"
+        - "service"
+        - "region"
+        - "version"
+      tag_values:
+        - "prod"
+        - "staging"
+        - "dev"
+        - "pod-a"
+        - "pod-b"
+        - "pod-c"
+        - "web"
+        - "api"
+        - "worker"
+        - "us-east-1"
+        - "us-west-2"
+        - "eu-west-1"
+        - "v1"
+        - "v2"
+        - "v3"
+      name_length:
+        inclusive:
+          min: 1
+          max: 32
+      tag_length:
+        inclusive:
+          min: 3
+          max: 16
+      tags_per_msg:
+        inclusive:
+          min: 2
+          max: 5
+      unique_tag_ratio: 1.0
+      value:
+        float_probability: 0.5
+        range:
+          inclusive:
+            min: -9999999
+            max: 9999999
+      multivalue_count:
+        inclusive:
+          min: 2
+          max: 32
+      multivalue_pack_probability: 0.08
+      kind_weights:
+        metric: 100
+        event: 0
+        service_check: 0
+      metric_weights:
+        count: 0
+        gauge: 0
+        timer: 0
+        distribution: 1
+        set: 0
+        histogram: 0
diff --git a/test/correctness/dsd-tag-filterlist-post-aggr/config.yaml b/test/correctness/dsd-tag-filterlist-post-aggr/config.yaml
new file mode 100644
index 00000000000..3c5fa449481
--- /dev/null
+++ b/test/correctness/dsd-tag-filterlist-post-aggr/config.yaml
@@ -0,0 +1,19 @@
+analysis_mode: metrics
+millstone:
+  image: saluki-images/millstone:latest
+  config_path: millstone.yaml
+datadog_intake:
+  image: saluki-images/datadog-intake:latest
+  config_path: ../datadog-intake.yaml
+baseline:
+  image: registry.datadoghq.com/agent:7.76.2-jmx
+  files:
+    - datadog.yaml:/etc/datadog-agent/datadog.yaml
+  additional_env_vars:
+    - DD_API_KEY=correctness-test
+comparison:
+  image: registry.datadoghq.com/agent-dev:ca27cfe6-py3-jmx
+  files:
+    - datadog.yaml:/etc/datadog-agent/datadog.yaml
+  additional_env_vars:
+    - DD_API_KEY=correctness-test
diff --git a/test/correctness/dsd-tag-filterlist-post-aggr/datadog.yaml b/test/correctness/dsd-tag-filterlist-post-aggr/datadog.yaml
new file mode 100644
index 00000000000..3919762b6c8
--- /dev/null
+++ b/test/correctness/dsd-tag-filterlist-post-aggr/datadog.yaml
@@ -0,0 +1,48 @@
+# Using a fixed hostname ensures consistent tags between the baseline and comparison agents.
+hostname: "correctness-testing"
+
+# Dummy API key.
+api_key: dummy-api-key-correctness-testing
+
+# We have to specifically configure the health port to use.
+health_port: 5555
+
+# Point ourselves at the datadog-intake service.
+dd_url: "http://datadog-intake:2049"
+
+# Turn off UDP and listen on a UDS socket instead.
+dogstatsd_port: 0
+dogstatsd_socket: /airlock/metrics.sock
+
+dogstatsd_origin_detection: false
+
+# Keep worker count at 1 to ensure deterministic gauge processing.
+dogstatsd_workers_count: 1
+
+metric_tag_filterlist:
+  - metric_name: tagfilter.exclude
+    action: exclude
+    tags:
+      - pod
+  - metric_name: tagfilter.include
+    action: include
+    tags:
+      - env
+  - metric_name: tagfilter.exclude.multi
+    action: exclude
+    tags:
+      - pod
+      - region
+  - metric_name: tagfilter.include.multi
+    action: include
+    tags:
+      - env
+      - service
+  - metric_name: tagfilter.conflict
+    action: include
+    tags:
+      - env
+  - metric_name: tagfilter.conflict
+    action: exclude
+    tags:
+      - pod
diff --git a/test/correctness/dsd-tag-filterlist-post-aggr/millstone.yaml b/test/correctness/dsd-tag-filterlist-post-aggr/millstone.yaml
new file mode 100644
index 00000000000..835dbc78aca
--- /dev/null
+++ b/test/correctness/dsd-tag-filterlist-post-aggr/millstone.yaml
@@ -0,0 +1,74 @@
+seed: [2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61, 67, 71, 73, 79, 83, 89, 97, 101, 103, 107, 109, 113, 127, 131]
+target: "unixgram:///airlock/metrics.sock"
+aggregation_bucket_width_secs: 10
+volume: 20000
+corpus:
+  size: 20000
+  payload:
+    dogstatsd:
+      contexts:
+        constant: 1024
+      metric_names:
+        - "tagfilter.exclude"
+        - "tagfilter.include"
+        - "tagfilter.miss"
+        - "tagfilter.exclude.multi"
+        - "tagfilter.include.multi"
+        - "tagfilter.conflict"
+      tag_names:
+        - "env"
+        - "pod"
+        - "service"
+        - "region"
+        - "version"
+      tag_values:
+        - "prod"
+        - "staging"
+        - "dev"
+        - "pod-a"
+        - "pod-b"
+        - "pod-c"
+        - "web"
+        - "api"
+        - "worker"
+        - "us-east-1"
+        - "us-west-2"
+        - "eu-west-1"
+        - "v1"
+        - "v2"
+        - "v3"
+      name_length:
+        inclusive:
+          min: 1
+          max: 32
+      tag_length:
+        inclusive:
+          min: 3
+          max: 16
+      tags_per_msg:
+        inclusive:
+          min: 2
+          max: 5
+      unique_tag_ratio: 1.0
+      value:
+        float_probability: 0.5
+        range:
+          inclusive:
+            min: -9999999
+            max: 9999999
+      multivalue_count:
+        inclusive:
+          min: 2
+          max: 32
+      multivalue_pack_probability: 0.08
+      kind_weights:
+        metric: 100
+        event: 0
+        service_check: 0
+      metric_weights:
+        count: 0
+        gauge: 0
+        timer: 0
+        distribution: 1
+        set: 0
+        histogram: 0
diff --git a/test/correctness/dsd-tag-filterlist/config.yaml b/test/correctness/dsd-tag-filterlist/config.yaml
new file mode 100644
index 00000000000..5c90f3125ae
--- /dev/null
+++ b/test/correctness/dsd-tag-filterlist/config.yaml
@@ -0,0 +1,22 @@
+analysis_mode: metrics
+millstone:
+  image: saluki-images/millstone:latest
+  config_path: millstone.yaml
+datadog_intake:
+  image: saluki-images/datadog-intake:latest
+  config_path: ../datadog-intake.yaml
+baseline:
+  image: registry.datadoghq.com/agent:7.76.2-jmx
+  files:
+    - datadog.yaml:/etc/datadog-agent/datadog.yaml
+  additional_env_vars:
+    - DD_API_KEY=correctness-test
+comparison:
+  image: saluki-images/datadog-agent:testing-release
+  files:
+    - datadog.yaml:/etc/datadog-agent/datadog.yaml
+  additional_env_vars:
+    - DD_API_KEY=correctness-test
+    - DD_DATA_PLANE_ENABLED=true
+    - DD_DATA_PLANE_DOGSTATSD_ENABLED=true
+    - DD_AGGREGATE_CONTEXT_LIMIT=500000
diff --git a/test/correctness/dsd-tag-filterlist/datadog.yaml b/test/correctness/dsd-tag-filterlist/datadog.yaml
new file mode 100644
index 00000000000..e7fdbb9a72b
--- /dev/null
+++ b/test/correctness/dsd-tag-filterlist/datadog.yaml
@@ -0,0 +1,56 @@
+# Using a fixed hostname is both required to avoid errors, and also ensures
+# consistent tags between the baseline Agent and the ADP comparison target.
+hostname: "correctness-testing"
+
+# Dummy API key.
+api_key: dummy-api-key-correctness-testing
+
+# We have to specifically configure the health port to use.
+health_port: 5555
+
+# Point ourselves at the datadog-intake service.
+dd_url: "http://datadog-intake:2049"
+
+# Turn off UDP and listen on a UDS socket instead.
+dogstatsd_port: 0
+dogstatsd_socket: /airlock/metrics.sock
+
+# Keep origin detection enabled to stay closer to real DogStatsD behavior.
+#
+# Note: the current corpus still primarily exercises instrumented-tag filtering;
+# enabling origin detection did not change the final comparable metric set in the
+# latest parity run.
+dogstatsd_origin_detection: true
+
+# Gauges can be processed out-of-order when multiple workers are used, while ADP
+# does not use multiple workers. Keep worker count at 1 to remove that
+# difference from this parity test.
+dogstatsd_workers_count: 1
+
+metric_tag_filterlist:
+  - metric_name: tagfilter.exclude
+    action: exclude
+    tags:
+      - pod
+  - metric_name: tagfilter.include
+    action: include
+    tags:
+      - env
+  - metric_name: tagfilter.exclude.multi
+    action: exclude
+    tags:
+      - pod
+      - region
+  - metric_name: tagfilter.include.multi
+    action: include
+    tags:
+      - env
+      - service
+  - metric_name: tagfilter.conflict
+    action: include
+    tags:
+      - env
+  - metric_name: tagfilter.conflict
+    action: exclude
+    tags:
+      - pod
diff --git a/test/correctness/dsd-tag-filterlist/millstone.yaml b/test/correctness/dsd-tag-filterlist/millstone.yaml
new file mode 100644
index 00000000000..835dbc78aca
--- /dev/null
+++ b/test/correctness/dsd-tag-filterlist/millstone.yaml
@@ -0,0 +1,74 @@
+seed: [2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61, 67, 71, 73, 79, 83, 89, 97, 101, 103, 107, 109, 113, 127, 131]
+target: "unixgram:///airlock/metrics.sock"
+aggregation_bucket_width_secs: 10
+volume: 20000
+corpus:
+  size: 20000
+  payload:
+    dogstatsd:
+      contexts:
+        constant: 1024
+      metric_names:
+        - "tagfilter.exclude"
+        - "tagfilter.include"
+        - "tagfilter.miss"
+        - "tagfilter.exclude.multi"
+        - "tagfilter.include.multi"
+        - "tagfilter.conflict"
+      tag_names:
+        - "env"
+        - "pod"
+        - "service"
+        - "region"
+        - "version"
+      tag_values:
+        - "prod"
+        - "staging"
+        - "dev"
+        - "pod-a"
+        - "pod-b"
+        - "pod-c"
+        - "web"
+        - "api"
+        - "worker"
+        - "us-east-1"
+        - "us-west-2"
+        - "eu-west-1"
+        - "v1"
+        - "v2"
+        - "v3"
+      name_length:
+        inclusive:
+          min: 1
+          max: 32
+      tag_length:
+        inclusive:
+          min: 3
+          max: 16
+      tags_per_msg:
+        inclusive:
+          min: 2
+          max: 5
+      unique_tag_ratio: 1.0
+      value:
+        float_probability: 0.5
+        range:
+          inclusive:
+            min: -9999999
+            max: 9999999
+      multivalue_count:
+        inclusive:
+          min: 2
+          max: 32
+      multivalue_pack_probability: 0.08
+      kind_weights:
+        metric: 100
+        event: 0
+        service_check: 0
+      metric_weights:
+        count: 0
+        gauge: 0
+        timer: 0
+        distribution: 1
+        set: 0
+        histogram: 0