Recently, during a production incident response, I guessed the root cause of an outage correctly within less than an hour (cool!) and submitted a fix just to rule it out, only to then spend many hours...