Moved issue gerrit:5225 to now be  issue chromium:678698 .

I have no idea how or why this would be different -- both Gerrit and Rietveld use the same CQ Dry Run code.

Can you point me to an example of this behavior? Best would be the same CL uploaded to both systems with different trybot behavior; but any pair of random CLs exhibiting this would be okay as well.

See https://pdfium-review.googlesource.com/c/2153/2 for example. It is retrying the try runs for the win bots.

Can you please point me at a pdfium CL on rietveld where some dryrun jobs failed and didn't retry? I'm still not sure that this is a regression.

Oh, I think you're right. I just never noticed on Rietveld because they get hidden behind the more>> link where as it's obvious in Gerrit that these retried.

Feel free to close as user error.

Ah that makes sense. We can look into being more aggressive with hiding old failed jobs in the Gerrit UI. Thanks!

You can change your cq.cfg to tell it to not retry failures: https://cs.corp.google.com/eureka_internal/chromium/tools/depot_tools/third_party/cq_client/cq.proto?l=189 

specifically, "try_job_retry_quota 0" should do the trick.

In fact, it's great if your tests are so good you don't need to retry. We might want to consider making no retries a default in CQ.