2.5.73-mm1 is fine.

This is *not* the "clock runs really really fas"t issue - I left -mm2 running overnight and
in some 8 hours the system clock only drifted a few seconds versus wall clock (and it's
possible it was off a few seconds when it booted, as it didn't get an NTP sync at boot).

Audio plays "too fast" - a 4 minute .ogg goes through in about 3:40, sounding a bit
high-pitched in the process.

lspci -v:
00:1f.5 Multimedia audio controller: Intel Corp. 82801CA/CAM AC'97 Audio (rev 02)
        Subsystem: Cirrus Logic: Unknown device 5959
        Flags: bus master, medium devsel, latency 0, IRQ 11
        I/O ports at d800 [size=256]
        I/O ports at dc80 [size=64]

relevant dmesg output:
Advanced Linux Sound Architecture Driver Version 0.9.4 (Mon Jun 09 12:01:18 2003 UTC).
request_module: failed /sbin/modprobe -- snd-card-0. error = -16
PCI: Setting latency timer of device 0000:00:1f.5 to 64
intel8x0: clocking to 51084
ALSA sound/pci/intel8x0.c:2520: joystick(s) found
ALSA device list:
  #0: Intel 82801CA-ICH3 at 0xd800, irq 11

The 'clocking to 51084' is *VERY* suspicious, as previously this value was
*always* 48000. Something very strange obviously happened in
intel8x0_measure_ac97_clock(), but I can't figure out what.  I don't
think the mdelay(50) is off - the Bogomips value hasn't changed from 3185.04.
The problem is deterministic - on 3 reboots, I've gotten 51084 twice and 51085
once. Unless an odd latency is hitting spin_lock_irq(save,restore), I don't
see how that code can break?

Any ideas?