"database is full" with TMPDIR pointing at a 5TB partition: 3̶2̶-̶b̶i̶t̶ ̶b̶u̶g̶?

Woaah. So so far, “pragma quick_check” (!) has taken seven hours, of which 9 minutes was actual CPU time. It’s become increasingly depressing how a single backup run with only minor changes to 1.4TB of data is taking 24-48h to complete; I wonder is there anything I can do to optimise database performance on this front? It’s on some pretty good 7200rpm drives, and the database is 24GB.

Wondering whether adding the index described here might help; will give a try once the integrity check completes.