We have some validate unit tests that are rarely failing to find data corruption. The entries logged after the validate cmd are showing the data corruption that should have been found. The next step is adding logging of the collection and index table entries that validate sees as validate sees the entries. These log messages can be specially toggled on for unit tests only.

Validate is failing to see any data corruption after the first lighter-weight round of validation. There's a second round that collects more information if the first round finds corruption. The test failures show that only the first round runs, so nothing is found in that first round. We need logging for the first round of validation, to narrow down where the error takes place in the code. The first round of validation uses buckets of counters, incrementing for collection table entries and then decrementing for index table entries.