Depending on what caused the error, retries might not help. The outage seemed to repro consistently for certain systems, and for quite a while (an hour?). If a DNS issue, a client machine might keep retrying with the cached IP it got from an earlier DNS request.

Depending on what caused the error, retries might not help. The outage seemed to repro consistently for certain systems, and for quite a while (an hour?). If a DNS issue, a client machine might keep retrying with the cached IP it got from an earlier DNS request.

Yeah, but in this particular incident, I was seeing the failures be random, so retrying fixed it if enough retries were done.

I think it's too many retries. This is supposed to be about the developer's connection being bad, not about (rare) CDN incidents. We won't print out the "go get the offline package" message for a really long time for these folks now. This adds 2.5 minutes, right?

@bobbrow I reduced the max delay between retries to 15 seconds: #7883 .