Retry internal signals to make sure events don't hang

We've seen some issues where a cherry-pick process inexplicably hangs, but restarts correctly when the bot is next restarted. This would indicate that some internal event emitters are not being received correctly, but there is currently no internal architecture to ensure that a signal is received by the application.

This should also extend to internal event listeners. Possible solution is an event cache where events are stored until they are consumed or expire.