proc: table is full & process detected outside of its chroot

[Thomas Klausner <wiz%NetBSD.org@localhost>]

Hi!

On my main server, something weird happened this night.
First, from around 0:52 on, there was network trouble like

       dhcpcd[1599]: igc0: xxxx:redacted:xxxx is unreachable

for some hours.

Around 4:22 I see the first

       proc: table is full - increase kern.maxproc or NPROC

(these keep coming) followed by, around 4:36 to 4:38, many of these:

       chrooted pid 765 uid 2000 (cc1) detected outside of its chroot

around 4:40 I start seeing

       inetd[5364]: fork: Resource temporarily unavailable

In the morning, I couldn't even login, and shutting down via the power
button didn't work either because powerd couldn't start the shutdown
script.

I have mostly two questions:

1. Could there really have been a chroot escape, or is the detection
   logic unreliable when the process table is full?

2. How can we improve shutdown via power button in such situations? Or
   how should I have shut down the machine?

Thanks,
 Thomas