I am occasionally getting an read ECONNRESET while reading via a pipeline, during GitHub actions that need to download binary archives with tools, to satisfy dependencies (via xpm & xPacks).

Currently I'm manually restarting the failed jobs, but I was thinking perhaps implementing a small loop in xpm, with several retries, might fix these issues.

Did anyone try such solutions? What would be the best strategy? Is there any way to reuse the connection, or a new fetch() is required? Is there a recommended delay between retries, to avoid upsetting the GitHub server with too many requests?