After we add metrics around delinquency, we should use our existing tests to determine how often delinquent operations occur. This could involve:

• Running our test suite with a server binary configured to crash any time a delinquent operation is reported
• Introducing new test suites that fail when an operation is delinquent, and report information about it

If we find delinquency is associated with certain operations, we should file tickets about them. At the same time, we should make sure the new diagnostics are not overly noisy, and that they actually provide useful information.