publish_tarball running twice for 95.0.4638.69

[Evangelos Foutras]

I don't think I've ever seen this happen before but there appear to be two jobs for 95.0.4638.69 running concurrently:

• https://ci.chromium.org/p/infra/builders/cron/publish_tarball/b8832079842152008545
• https://ci.chromium.org/p/infra/builders/cron/publish_tarball/b8832072929834332097

I also can't recall a time before today where more than one job was active at the same time.

I think I'll wait for the second job's export_tarball step to finish to make sure I don't build Chromium from a tarball that's going to be overwritten and invalidate its checksum.

[Kubo Da Costa, Raphael]

Well this is interesting.

https://ci.chromium.org/ui/p/infra/builders/cron/publish_tarball_dispatcher/b8832092730639116273/overview seems to have triggered the first run, while https://ci.chromium.org/ui/p/infra/builders/cron/publish_tarball_dispatcher/b8832081407440285857/overview triggered the second one. IIUC the dispatcher code just checks if a tarball corresponding to a given version exists in Google Storage or not and, if the bot creating the tarball hasn't finished running yet, the dispatcher will trigger another run and we end up in this situation. And I _think_ the one that finishes running last overwrites the previous tarball.

I'm not really sure of whether there's a way to stop the bot from even doing the second run or if we should let it work but just not run gsutils upload if the tarball's already present.