Since we're now building all our images in separate parallel workflow runs, we might be able to just stagger them all by a random number of seconds up to a minute or two, just so we don't stampede in the first few minutes when the workflow is kicked off.
Description
Release workflows fail somewhat regularly with rate-limiting errors:
(link)
The broader workflow run that hit this is here: https://github.com/chainguard-images/images/actions/runs/3830355184 -- it's got 55 workflows running in parallel, most of which seem to complete within 5-10 minutes, when they don't hit the limit.
Notes:
retry-after
above is actually not that much -- if we just waited 10ms, we wouldn't have hit the ceiling?Since we're now building all our images in separate parallel workflow runs, we might be able to just stagger them all by a random number of seconds up to a minute or two, just so we don't stampede in the first few minutes when the workflow is kicked off.