The Ticket That Started It
A query was taking 12 seconds pulling from a staging table that the morning SSIS package loads. Someone opened the execution plan, spotted a clustered index scan, and added a non-clustered index. The query dropped to 400ms. Ticket closed.
Three weeks later, the SSIS package started timing out. The ETL window that used to finish in 40 minutes was now running 90. Nobody connected the two events as they happened weeks apart and the symptoms looked completely unrelated. Different team members, different Jira boards, different oncall rotations.