At some point I seem to remember that we came up with a fall-back that PRs would eventually be opened (after a generous period) even in case of solver errors - to avoid the bot getting stuck indefinitely where a human might easily fix the recipe. I'm not 100% sure anymore if this extends to missing parents, but recently the bot seems to have started opening quite a few PRs despite those feedstocks having missing ancestors.
This was noted as surprising by a few users (e.g. here, here), and also affects our migration closures, which now happen even earlier (example), since the standard "all PRs have been opened" (rather than "all PRs have been merged") is basically satisfied instantly once missing parents are ignored.
At first I thought this was just about some very long-running migrations eventually reaching that timeout, but given that there have been 3 cases in less than 24h, I'm beginning to think this might be due to a recent change, e.g. #6440
CC @beckermr @isuruf
At some point I seem to remember that we came up with a fall-back that PRs would eventually be opened (after a generous period) even in case of solver errors - to avoid the bot getting stuck indefinitely where a human might easily fix the recipe. I'm not 100% sure anymore if this extends to missing parents, but recently the bot seems to have started opening quite a few PRs despite those feedstocks having missing ancestors.
This was noted as surprising by a few users (e.g. here, here), and also affects our migration closures, which now happen even earlier (example), since the standard "all PRs have been opened" (rather than "all PRs have been merged") is basically satisfied instantly once missing parents are ignored.
At first I thought this was just about some very long-running migrations eventually reaching that timeout, but given that there have been 3 cases in less than 24h, I'm beginning to think this might be due to a recent change, e.g. #6440
CC @beckermr @isuruf