Warning: This build has drifted.
The coverage report for this pull request build may be inaccurate because its base commit is no longer the HEAD of its target branch.
This means it includes changes from outside the original pull request, including, potentially, unrelated coverage changes.

    • Learn more: For more information on this, see Tracking coverage changes for pull request builds.
    • Fix now: For a quick fix, rebase this PR at GitHub. Your next report should be accurate.
    • Prevent going forward: To avoid this issue with future PRs, see these Recommended CI Configurations.

fix(cacher): restore pre-regression test jitter envelope; sync test on cacher

Follow-up to d17d67386 (cherry-picked from #3370). That fix set the
cacher's reschedule delay to zero in :test, which is effectively a hot
reschedule loop — each :do_cache handler completion immediately enqueues
the next, burning scheduler cycles across every cacher in the suite.

Root cause is upstream: b6da32ecb (Oct 2025) replaced the predecessor
RecentInsertsBroadcaster's `max(@cadence, :rand.uniform(cluster_size * 200))`
formula — which collapsed to a [100, 200] ms envelope on single-node test
runs — with an unconditional `:rand.uniform(1_000)`, pushing test delays
to [101, 1100] ms and regressing any test polling on Counters updates.

Gate the jitter ceiling on :env (100 in test, 1_000 in prod) so the test
envelope is back in pre-refactor range while prod behavior is unchanged.

Also drop `retry_assert` in test/logflare/backends_test.exs's "caches
latest timestamp correctly" test in favor of an explicit
`send(cacher, :do_cache) + :sys.get_state/1` fence — the test no longer
depends on whatever the cacher's timer happens to be, so future tuning
of the interval won't silently re-tripwire it. Test runs in ~1 ms
instead of ~100 ms.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>