Bug in anti join conditions of Miovision `identify_zero_counts

[gabrielwol]

Anti join conditions don't account for case when there is an existing null classification a_r and a new, non-null a_r to be inserted. This line needs an OR existing.classification_uid IS NULL condition:

AND COALESCE(existing.classification_uid, 0) = COALESCE(new_gaps.classification_uid, 0)

@gabrielwol Actually, these short little blips created for classification_uid=10 take place during a larger anomaly, (suspect, questionable).

These are a different problem-level, do-not-use, and investigation level though. I guess I don't see a way that I would have done it much differently in an automated way... Some manual fixing needed here though.

Originally posted by @Nate-Wessel in https://github.com/Toronto-Big-Data-Innovation-Team/monitoring/pull/60#discussion_r1478904965