Sorry, typing on a handheld with too thick thumbs :)
The database files can be on remote storage (e.g. Windows server or NAS) (I know that this may be a factor and we discussed this already and I know the texts on your web site about false/incomplete locking etc). I just mentioned that here to show that using pragma quick_check is not practicable in a production environment when you deal with larger databases. All recent occurrences of the problem where on local hard disks / SSD disks. It even affects the small configuration databases my application maintains in SQLite, with the FULL synch mode for maximum safety. These also use WAL / shared_cache but are only updated a couple of times per minute and have to endure much less stress by concurrent multi-thread read/write.