Compaq Armada 1500c APM weirdness

[Ian Smith]

I've scoured the -mobile archives, finding no mention of this machine at
all.  If this is (an old) FAQ please point me in the right direction. 
Sorry if it's a bit long ..

I've had this beastie for about 4 months; Celeron 300, 4Gb HD, currently
64Mb.  The first week I shrunk its installed Win95b partition and
installed 4.5-RELEASE and KDE 2.2 in 2.5Gb as a testbed to see if it was
worthwhile fitting a 20Gb drive and 160Mb.  I then installed Win98se
plus all of Compaq's additional 1500c bits, upgraded setup/diagnostics
and flashed the latest BIOS.  All without incident, or any ill effect on
how well FreeBSD ran.

After nearly a month's uptime after building a kernel that got sound
working, during which time I'd been carting it around, using it lots on
battery with short APM timings for screen / HD / suspend BIOS settings
of 1/2/3 minutes, working just fine in or out of X, I had to shut it
down and boot Win98 for a brief job unloading a Kodak DC3200 camera
(serial).

Ever since then, APM auto-suspend has refused to work in FreeBSD. 
Screen and HD timeouts still work, suspend button and zzz work, battery
low suspend works, everything is peachy except auto-suspend, which won't
work with or without X running.  I should mention that I always had to
go via setup after running Win98 before rebooting FreeBSD, as Win98
zeros these timeouts for its own use, but resetting desired timeouts had
always worked before.

ACPI is now set off, but made no difference before when it was on those
first three months, presumably ignored.  My gut feeling is that Win98,
apart from clearing these timeouts as mentioned, may have toggled
something else affecting APM, but nothing that shows in the BIOS
settings?  I've tried all manner of fiddling with the Win98 APM settings
too; no difference. 

Does this ring any bells for anyone?

Cheers, Ian