When Hoff fails to promote a passing build due to a new commit on master for example, it should restart the testing build of all PRs in the train. Currently it only restarts the first immediately. The others end up waiting a while for a restart.

Here's the original bug report:

On two PRs in a merge train, Hoff has restarted jobs after their builds passed. This should not happen. (cf. the two PRs that link here, this happened between 10:00 and 10:30)

Timeline

• @ 10:00:23 - merge order on #524
• @ 10:00:38 - "rebased as" comment on #524
• @ 10:00:42 - job started on branch testing/524 (f9928ad)
• @ 10:05:## - unknown to us at the time a new commit 542c613 is pushed into master!
• @ 10:06:48 - merge and tag order on #329
• @ 10:07:05 - "speculatively rebased as" comment on #329
• @ 10:07:09 - job started on branch testing/329 (59193b1)
• @ 10:10:26 - build succeeded on f9928ad, i.e. testing/524
• @ 10:10:32 - promotion fails, git push failed -- reason: [rejected] f9928ad -> master (fetch first) (outdated testing branch!)
• @ 10:10:32 - hoff decides to try to integrate again
• @ 10:10:47 - posted comment "rebased as 3eedaf4" on #524
• @ 10:10:## - Hoff should have restarted #329 as it is based on top of #524 but it dit not!
• @ 10:20 - #524 is finally merged
• @ 10:22 - rebased as 0ba5eb3 on #329
• @ 10:37 - #329 is finally merged