For what it's worth, I have an alarm clock connected to the French grid, and until last month it's been perfectly on time, so assuming it is tuned to 50Hz there is probably something wrong with the source data.
Agreed - without knowing more about the data provenance, all I can show here are deviations from the long-term trend. It's enough to see the recent problem clearly, but I wouldn't trust the increase after day 900 to be real. Most likely something changed in how they report the data around that time.
