[deverman]

Summary

• capture timeout queue and process state immediately, before passive cooldown
• remove active Bridge health diagnostics and the hidden recovery-delay environment behavior
• replace split cooldown/recovery waits with explicit 13-second smoke/stress defaults and 15-second release waits
• preserve production transport, query behavior, BridgeClient recovery, and the 45-second deadline

Closes #222

Acceptance journey

A release benchmark timeout immediately records the untouched IPC queue/process state, waits 15 seconds without sending diagnostic Bridge requests, and leaves enough metadata to distinguish a late response from a stranded request.

Validation impact

performance

Validation

• swift test
• bash -n scripts/benchmark-suite.sh
• git diff --check origin/master...HEAD
• swift run focusrelay-dev validate --impact performance in a clean live session
• smoke profile
• release retry only if smoke is clean

The branch is intentionally draft until live validation completes. Current master includes the Forecast production change from #224, so any eventual release evidence applies to that newer production fingerprint, not the August candidate.

[deverman]

Local validation after rebasing onto current master (77995e9):

• swift test: passed, including the passive timeout artifact contract and the new Forecast suites
• bash -n scripts/benchmark-suite.sh: passed
• git diff --check origin/master...HEAD: passed
• branch is clean and synchronized with origin/perf/222-passive-timeout-diagnostics

Live performance validation and smoke are intentionally pending. The machine currently has a persistent unrelated appstoreagent workload at ~86% CPU, and pmset -g batt reports the battery discharging despite AC power. Starting the benchmark under those conditions would violate #222’s clean-session acceptance criteria. No plugin install/restart or benchmark was started.

[deverman]

Validation update for commit eb7525d:

• swift run focusrelay-dev validate --impact performance: passed, including the passive timeout-artifact contract and live semantic gates.
• Smoke profile: passed at .build/benchmarks/smoke-20261018-011342; 255/255 measured calls succeeded, 0 errors, 0 timeouts, and all timeout diagnostic files were empty.
• One release retry under caffeinate: passed at .build/benchmarks/release-20261018-121937; 745/745 measured calls succeeded, 0 errors, 0 timeouts, complete scenario coverage, and the suite evidence checker passed.
  • task counts p95: 3294.82 ms
  • list tasks slowest scenario p95: 14485.43 ms
  • project counts p95: 15067.01 ms
• Branch is based on current origin/master (77995e9), is one commit ahead, working tree clean, and CI is green.

Release blocker: the formal production fingerprint changed from 31c87ccc228cbb31f2690e27a46d763c4c98d67d49118c559206315dd0af69a2 on origin/master to 43fff751265435eb4e7e2814c6f8ef5c1c7e4368a89e337bac546b3ac14dfabb on this branch. The plugin is unchanged, but the benchmark files are under the shipped FocusRelayCLI target and the fingerprint includes all of Sources. Per the go/no-go rule, this candidate is not being marked release-ready and the PR remains draft.

[deverman]

Release decision update:

The production fingerprint change is explicitly accepted. It is caused by passive benchmark tooling living in the shipped FocusRelayCLI target; the Bridge plugin, transport behavior, 45-second deadline, and recovery branches are unchanged.

The branch was rebased onto current master at 52a14f0. That intervening commit changes registry packaging/docs only and leaves the validated production fingerprint inputs unchanged. Rebased candidate: 76d05df.

Validation retained:

• performance validation passed
• smoke passed with 255/255 successful measured calls
• release profile passed with 745/745 successful measured calls and zero timeouts
• full swift test passed again after rebase

The earlier fingerprint blocker is closed by explicit owner acceptance. No tag, merge, or publication is implied by this decision.