For MediaWiki (recent comments | status changes | tags | authors | states | release notes | statistics)
How about failing fast with an exception?
Well, that's what the existing behavior was. It was quite intrusive on the cluster when we deployed 1.17.
So did you find the cause?
No. If we had looked at the backtraces back then, we might have.
How about we put in a logging thing that logs backtraces for these things but doesn't throw user-visible error messages?
That's fine with me. I haven't seen these issues myself. It concerns me if we are hiding likely bugs.