Is there a way to extract statistics about repo repair process? The average / total number of stalls, per job type maybe, per slave? Main reason for stalls, that sort of thing?
Mmmm, unfortunately we don’t really log those anywhere currently, outside of the application logs which would be a nightmare to parse out I think.
Might be a good thing to track as aggregated statistics though, as you say. Shouldn’t be too bad to do either, since all that code is fairly centralized. I’ll add it to the wishlist 
Cheers,
Jon