Would it be possible to some delay between retries? It seems like webservers are not restarted atomically and retries /wo delay fail all. We run hundreds pipelines per hour and even a percentually small failures are a significant number per day.
Like 2nd retry /wo delay, 50 ms for 3rd retry, 100 ms for 4th retry, ... and retry by default at least 8x (eg. 50, 100, 200, 400, 800, 1600 ms = about 3 seconds in total)