Keep track of the number of commands that exhaust their retry limit.

While we print failure messages on the console, sometimes logs are lost or
overwhelmed. Keeping a count of how many times we've failed retriable commands
helps get a magnitude of the problem.