Yep, I’ve been seeing that, too. Although, I think I’ve only seen it on an initial search. Going to page 2, or “back,” I don’t think I’ve ever gotten the error in those cases. Probably just lucky.
Happens all the time to me. Actually, seach/lookup related features work 50% of the time for me which is really something. I understand the need for rate limiting, but this is not the way. Even most userscripts I have will retry until result is there, unlike MB, which requires from user to fix software shortcoming.
As an example I constantly get retry on release group when adding a new release and when searching for track artist.
We fixed this long standing issue with the search: one of our MB servers was unable to access to the search server, due a misconfiguration in an ACL.
So, depending on which server the search was originating from, the search cluster was returning an error, and when retrying most of the times it was done from another node, so the query was working, giving the impression of randomness.
Amazing. Sometimes I feel bad about complaining about these minor inconveniences, but it sounds like the relevant person didn’t even know there was an issue?