The autosave interval should be as short as possible, so that ideally everything could be recovered if (when) needed. The shortest interval the GUI lets me specify is one minute, I'd rather went with 30 s, but 15 minutes is just way too much.
The kernel caches the reads and writes anyway, and if disk activity is of a greater concern than the possibility of recovery, one should adjust those cache timeouts rather then the ooo interval anyway.
Probably is an upstream issue, so please forward if appropriate.
Please report this upstream, there is nothing we can do here: