Signed build queue should have an automatic "healing" process

[RikkiGibson]

When the signed build goes down for some arbitrary reason it means we lose RPS data for changes introduced while the queue is down.

It's possible this is not tractible to solve in general, because often we have to merge some tactical fix in order to unblock the queue, but it would be so useful to be able to reliably re-queue a set of failed builds in order to be able to fill in the missing data.

At the least, having some documented procedure/guidelines for how to recover and get data in this situation would probably help make it feel less daunting :). This could probably apply both to cases where the signed queue went down temporarily or just cases where we have to play PR Detective and figure out which change introduced a regression.