I finally got to try out the kernel I built Tuesday afternoon, and
it has some significant problems compared to the one I have from
the 10th.

In particular, the Jan10 kernel would report that it could not ping
a cpu no more often than every couple of minutes, and sometimes
not for a few hours.  The Jan13 kernel, on the other hand, reported
about a hundred such logs between each line of the regular boot
messages, finally ending in a panic about "simple_lock: locking
against myself", before allowing a login.

If anyone wants more information about the exact messages, I'll
get them back to you within a day or so...  So, was the new breakage
temporary, and I should just rebuild tomorrow? Or do I need to get
a crashdump and play with gdb?  On the whole, it's been remarkably
stable for me though, two or three days uptime easily, with a
rebuild of userland in there somewhere.

Jon