Hey,

Looking at the statistics page I was wondering why the rate of change around the end of month shows sudden spikes:
https://musicbrainz.org/statistics/timeline/main#r+g/1495997787551.055/1542585600000+-recording

Issue: end of months, especially February (28 or 29 days) shows an increase in change, while the first days are usually a bit 'slower'. This seems unrealistic and is also not reflected in the actual data (eg core entities based on numbers).

Might this be caused by the incorrect dealing of month lengths in the calculation? Is there a 4 week based system?
Rate of change based on days should be used instead.

Attaching example screenshot.