Huy Do

The first and second AI make sense to me, and could be implemented easily. The last one, however, needs more clarification I think. Devs would likely force merge the PR...

> Do legitimate flaky failures tend to have generic error captures like `##[error]Process completed with exit code 1.`? Yup, that one and `Command docker exec -t ...` from Nova, i.e....

> +1 same thing for backward compat job! I have a fix ready here https://github.com/pytorch/test-infra/pull/5106, although it's weird that I don't see @ZainRizvi comment about lint on the issue, although...

@pytorchbot revert -m 'Sorry for reverting your change but I think there is a land race with the change https://hud.pytorch.org/pytorch/pytorch/commit/33e6791645b5950b0f39301f55b8a4a79c0ca847' -c landrace

Yes, please do a rebase to main and try to land this again

The stack is probably having a landrace with this change https://github.com/pytorch/pytorch/pull/122832 that was landed few hours ago

Looking at D51738439, it looks like @PaliC is already on this

@izaitsevfb raises a potential blocker where only github-export-checks has special security permissions to access the internet, i.e. calling GitHub

I guess we need to stop using Suo's token anyway, that's not a good practice to keep. So we can do the ~~first~~ second point and observe to see if...

@pytorchbot drci