After splitting the revision tables and migrating to a dedicated db, reviewing the training output is a good idea.
For example a false positive report should be possible to feed back into core and verify it is still a false positive.
Additionally a minimal WPEditSet should be tested (e.g. sampled edits) to ensure correctness for training.
There is some logic that was removed from trainer that can do this.
After splitting the revision tables and migrating to a dedicated db, reviewing the training output is a good idea.
For example a false positive report should be possible to feed back into core and verify it is still a false positive.
Additionally a minimal WPEditSet should be tested (e.g. sampled edits) to ensure correctness for training.
There is some logic that was removed from trainer that can do this.