"Judging from the spikes every decade prior to about 1800, it seems that Wikipedia generously applies birth year categories when birth decades would be more appropriate. Nonetheless, the birth categories turns out to be far better curated than the machine-readable PersonData field."

Looks like it's a Wikipedia problem.

I'd expect the prorating to spread the dates evenly across the decade precisely to avoid this kind of periodicity. It seems to suggest dates have been inaccurately recorded with year precision as, say 1690 when some author along the line didn't know if it was 1686 or 1694.