Wait a bit longer when retrying MergeOp after LOCK_FAILURE
In our environment, the race window for ref updates is surprisingly
long. Give it 15 seconds before retrying.
Remove the random component of the restart time. That sort of thing is
intended to prevent multiple instances moving in lockstep, but the
retries only happen on the instance that originally saw the lock
failure (i.e. received the original post to submit the change).
Change-Id: Ic890df08a873f736472127b0dcfbd78e5f574d69
1 file changed
