yep, as long as its skippable just like the vacuum is.
Such semi-transient events are hard to graps, or detect.
A Hard-crash during disk-IO just plain sucks.
I’d assume the only means of detecting are some time-constraints during DB scan/read.
Stuff like this shouldnt be happening on a journaled (ext3) fs though, afaik. I deffo wouldnt like the mt-daapd startup even further delayed on my slug.