[PerfectSlayer]

What Does This Do

Restores, through the bridge, the agent-side behaviors of the removed Feature Flags writers:

• Event delivery (ports of Harden Agentless Feature Flags EVP delivery #12477 and Send Java SDK identity with Feature Flagging EVP events #12575): each event batch is sent once; the agentless proxy endpoint is discovered once, recovered at most once a minute, and used only when the Agent forwards the DD-EVP-ORIGIN / DD-EVP-ORIGIN-VERSION identity headers sent with every event, through the proxy and direct intake alike.
• Shutdown flush: the SDK flushes its pending events through the agent shutdown: the tracing shutdown path when it runs, and otherwise (tracing or its shutdown instrumentation disabled) a Feature Flagging JVM hook registered once the SDK subscribes.
• Settings: declares the SDK settings the agent exposes (evaluation telemetry opt-out, exposures Datadog logging; ports of Add per-subject evaluation telemetry opt-out #12884 and Send feature flag exposures after evaluation, and let applications record them with ExposureHook #12774).
• Remote Configuration: a malformed flag configuration document is reported to Remote Configuration as not applied, instead of acknowledged, and closing Feature Flags releases its poller reference again.

Motivation

Keeps the behavior master has today with the agent-coupled writers (APMLP-1512).

Additional Notes

Files

File	Notes
lib EventProxy (+ test)	Send-once clients, discovery requiring the identity headers, recovery interval, definitive rejection handling
SDK DirectIntakeTransport (+ test)	Identity headers on direct intake
bridge AgentShutdownSource, JavaAgentBridge.shutdown()	Shutdown subscription through the gateway Backend.subscribeShutdown
lib FeatureFlagsBackend, agent FeatureFlaggingSystem (+ tests)	Shutdown listeners, lazy JVM hook (always registered once the SDK subscribes, idempotent with the tracing shutdown path), identity headers on the proxy clients, declared settings
dd-trace-api FeatureFlaggingConfig, metadata/supported-configurations.json	Declared SDK settings
AbstractOpenFeatureShutdownSmokeTest + 2 smoke tests	Pending flag evaluations flushed at application exit, with and without tracing, carrying the identity headers
AbstractSmokeTest	Mock Agent advertises the identity headers, like the Agent
deleted ShutdownInstrumentationTest	Feature Flags shutdown testing moved to the smoke tests
bridge JavaAgentBridge (+ DDOpenFeatureIntegrationDisabledForkedTest), dd-openfeature and dd-openfeature-1.69 builds	Runtime SPI check falling back to the SDK bridge if a helper misses an SPI method, to drop once dd-openfeature 1.69.0 is released (the japicmpBridgeV1 check then covers it); TODO APMLP-1512 markers for the 1.69.0 release, including the prepared muzzle pass directive
SDK RemoteConfigurationService, lib RemoteConfigServiceImpl (+ tests), bridge.v1 ConfigurationSource / gateway Backend Javadoc	The SDK rejects a malformed document by throwing; the agent rethrows the first subscriber failure once every subscriber is notified, so the poller reports an apply error; RemoteConfigServiceImpl.close() stops the reference-counted poller it started

Stack

The full change is in #12709. It was split into this stack to ease the review.

1. Resolve Feature Flags enablement from instrumenter config #12845
2. Add dd-openfeature SDK flag configuration and evaluation core #12840
3. Add dd-openfeature SDK settings and configuration delivery #12841
4. Add dd-openfeature SDK exposure and flag evaluation pipelines #12842
5. Add dd-openfeature SDK OpenFeature provider #12843
6. Port master Feature Flags changes to the dd-openfeature SDK #12957
7. Add the agent bridge for the dd-openfeature SDK #12975
8. Connect the dd-openfeature SDK to the agent #12844
9. Keep released dd-openfeature SDKs working with the agent #12976
10. Resolve the Feature Flags settings like the SDK #12977
11. Restore the agent Feature Flags behaviors through the bridge #12978 👈 this PR
12. Add agentless and standalone OpenFeature smoke tests #12846

Jira ticket: APMLP-1512

🤖 Generated with Claude Code

[datadog-datadog-prod-us1]

🎯 Code Coverage (details)
• Patch Coverage: 93.06%
• Overall Coverage: 59.18% (-0.35%)

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: de707ac | Docs | Give us feedback!}

[dd-octo-sts]

🟢 Java Benchmark SLOs — All performance SLOs passed

Suite	Status
Startup	🟢 pass

SLO thresholds are defined here based on automatically generated metrics. A warning is raised when results are within 5% of the threshold.

PR vs. master results

Scenario	Candidate	master	Δ (95% CI of mean)
startup:insecure-bank:iast:Agent	14.76 s	14.67 s	[-0.2%; +1.5%] (no difference)
startup:insecure-bank:tracing:Agent	13.56 s	13.61 s	[-1.1%; +0.3%] (no difference)
startup:petclinic:appsec:Agent	17.69 s	17.48 s	[+0.4%; +2.1%] (maybe worse)
startup:petclinic:iast:Agent	17.47 s	17.70 s	[-2.3%; -0.3%] (maybe better)
startup:petclinic:profiling:Agent	17.36 s	17.24 s	[-0.7%; +2.0%] (no difference)
startup:petclinic:sca:Agent	17.13 s	17.62 s	[-7.0%; +1.5%] (no difference)
startup:petclinic:tracing:Agent	16.42 s	16.35 s	[-5.4%; +6.3%] (unstable)

Commit: de707acb · CI Pipeline · Benchmarking Platform UI

---

Load and DaCapo benchmarks can be triggered manually in the GitLab pipeline. Results will appear in the Benchmarking Platform UI after completion.

[AlexeyKuznetsov-DD]

LGTM, posted 2 minor comments I found with Codex