See https://treeherder.mozilla.org/#/jobs?repo=mozilla-central&resultStatus=success%2Ctestfailed%2Cbusted%2Cexception%2Cretry%2Cusercancel%2Crunnable&group_state=expanded&selectedJob=239037152&revision=a1eb490ba4480b756fcc4f92f8dbc753c66d7900&searchStr=build%2Cship

The shippable builds are scheduled twice (for both normal push jobs and nightlies) if the gecko decision task for former not is complete when the cron job for latter starts.

What behavior would you expect to see in that case? We could error out if we don't find a decision task. I was originally concerned about what would happen when a cron task runs against a DONTBUILD commit, but those now have a decision task. I am inclined to not try polling waiting for the task to complete.

Ideally, check for the existence of a successful Gecko decision task. If there is none, sleep 5 minutes and try again. Retry 3 times.

Else sheriffs have to wait 10-15 minutes after a backout on central until they can request new nightlies (and not to focus that much on something else in the meanwhile that it gets delayed).

When a cron task depends on tasks from the on-push build, wait for the on-push
decision task to complete, and fail if it doesn't complete succesfully.

https://hg.mozilla.org/mozilla-central/rev/684d31a3f000