Number of days solving puzzles, II
dango alerted me to a new problem with the number of consecutive days solving puzzles count.
After some digging, I found the cause:
To make this computation more efficient, certain "hourly markers" are stored in the database.
Since August 19th, these markers are sometimes incorrect, I'm thinking because of the database
query that gets the last solved puzzle sometimes fails.
I found a possibly related bug: when someone is logged in and solves a puzzle, all relevant database tables
are locked (so read/writes don't conflict with each other). But this wasn't happening for solved puzzles
of people who are not logged in (!)
Fixed now.
I'll do a recompute of the consecutive days count, and will monitor this issue..