Merging this PR will not alter performance

✅ 11 untouched benchmarks

_{Comparing fix/flaky-response-aggregation-timings-multiple (417f97e) with master (692a2c4)}

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.76%. Comparing base (692a2c4) to head (417f97e).

@@           Coverage Diff           @@
##           master    #1737   +/-   ##
=======================================
  Coverage   99.76%   99.76%           
=======================================
  Files          33       33           
  Lines        3440     3440           
  Branches      473      473           
=======================================
  Hits         3432     3432           
  Misses          5        5           
  Partials        3        3           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

@bluetoothbot review

PR Review — test: widen safety margin in test_response_aggregation_timings_multiple

Targeted, well-reasoned fix for a Windows-CI flake. Math is internally consistent: the new pre-assertion gap below the 1020ms minimum widens from ~100ms to ~520ms (well clear of Windows' ~15.6ms default timer resolution), and the second sleep is rebalanced so the total elapsed time remains 1340ms — preserving the headroom of the assert-1-call branch exactly. The comment block is updated to match the new constants, the test still pins the same behaviour (RFC 6762 §14 protected aggregation: no send before 1020ms, exactly one send after the full window), and no production code is touched. Merge-ready.

Checklist

• Math/timing is consistent before and after the change
• Total elapsed time preserved (second-assert headroom unchanged)
• Comments updated to match new constants
• Test still asserts observable behaviour (send_mock call counts), not implementation
• No production code touched; scope matches PR description

Summary

Targeted, well-reasoned fix for a Windows-CI flake. Math is internally consistent: the new pre-assertion gap below the 1020ms minimum widens from ~100ms to ~520ms (well clear of Windows' ~15.6ms default timer resolution), and the second sleep is rebalanced so the total elapsed time remains 1340ms — preserving the headroom of the assert-1-call branch exactly. The comment block is updated to match the new constants, the test still pins the same behaviour (RFC 6762 §14 protected aggregation: no send before 1020ms, exactly one send after the full window), and no production code is touched. Merge-ready.

Automated review by Kōan417f97e