We have metrics like MediaWiki.centralauth.session.* and MediaWiki.session.* and MediaWiki.edit.failures.session_loss.rate which should be monitored for high failure or excess latency (the later are from new patches).
The last session problem (T102199) took a fair amount of user reports and debugging (including adding metrics) to try track down. At this point, it would be good to just have monitoring of these metrics to quickly detect these kind of problems, which cause all sorts of random "session lost" errors when people save edits (forcing re-submissions).