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.

executor: drain one old revision per reconcile

Preserves the observable "newest-first drain" invariant during
multi-revision rollouts (e.g. mid-rollout A→B→C where C is applied
before B finishes rolling out from A).

Before: scaleDownOld iterated all old revisions in one call, spending
budget across them. Fast clusters could drain B fully AND start
draining A within a single reconcile (or two back-to-back reconciles
within <100ms), making the "B drained to 0 while A > 0" transient
unobservable to 100ms polling.

After: once any revision receives drain calls in an iteration, break
out of the loop. The next reconcile picks up the next-newest revision
if this one is fully drained. Extra reconciles for multi-revision
cases; benefit is a cleanly observable drain order.

Verified live: e2e "Mid-rollout A→B→C (newest-first drain) — should
drain B (broken intermediate) before A (stable original)" trace now
shows the transient (A=6, B=0, C=8) before A begins draining.