A large share of my closed issues start life as a session replay or an error group. Null-pointer triage, binding errors, race conditions in dialogs, background jobs failing on bad data, a route provider going down. On the performance side: caching route lookups and pre-filtering candidates to speed up a booking flow, lazy-loading heavy tables, optimizing a bulk-download query, chasing CPU spikes. I also wrote the automation that turns monitoring alerts into GitHub issues so the loop runs itself.