TSTILE lockups on disk dump/backups

I=E2=80=99ve been seeing problems with my backups, which I do with dump, =
and pipe the data to gzip, and to an NFS-mounted filesystem.

I=E2=80=99ve had trouble with this from back in March of 2013. I posted =
to this list in June and July of 2013 when I was running MetBSD 6.0.2, =
also as recently as Feb 2019 when I was running NetBSD 7.2.

I thought this bug had been licked in 8.2, because I have not seen it in =
the past few years while running 8.2.

Unfortunately it=E2=80=99s back now that I have upgraded to NetBSD 9.4.

I thought that there was a bug report, but scanning in my email, I =
don=E2=80=99t see it. I did see chatter with the subject: "Re: tstile =
lockups - side note=E2=80=9D and =E2=80=9CRe: TSTILE lockups - not dead =
yet=E2=80=9D.

I=E2=80=99ve only been on 9.4 a month or two, and I=E2=80=99ve already =
seen the lockup at least 3 times. The dump processes get stuck on =
=E2=80=9Ctstile=E2=80=9D status, and can=E2=80=99t be killed. I have to =
reboot the system. If I wait too long, the whole system piles up on the =
tstiles and locks up hard.

Any thought about how I should deal with this? My kernel debugging =
skills are weak. Is there anything I can do to produce useful debug =
information next time this happens?

Many thanks,

-dgl-