Ok, this one has really gotten me, and google hasn't bailed me out yet, 
so I was hoping someone on the list could.

Some relevant system Info are an Athlon XP 1800+, nForce2 chipset 
motherboard, RT2561(?) wireless card, 40 GB internal (IDE) disk, and 
500GB External Western Digital MyBook USB drive, running MythBuntu 
7.10.  The kernel is 2.6.22-14-generic #1 SMP Tue Dec 18 08:02:57 UTC 
2007 i686 GNU/Linux.

Note that I (until recently) had the same setup/problems with a Fedora 
7/MythTV setup (on the same computer).

Relevant portion of /var/log/messages:
Feb 15 12:23:21 venus kernel: [23306.544344] usb 3-6: USB disconnect, 
address 3
Feb 15 12:23:25 venus kernel: [23310.022118] lost page write due to I/O 
error on dm-0
<repeat above ... a bunch of times>
Feb 15 12:23:33 venus kernel: [23318.120350] usb 3-6: new high speed USB 
device using ehci_hcd and address 4
Feb 15 12:23:33 venus kernel: [23318.254619] usb 3-6: configuration #1 
chosen from 1 choice
Feb 15 12:23:33 venus kernel: [23318.255811] scsi1 : SCSI emulation for 
USB Mass Storage devices
Feb 15 12:23:36 venus kernel: [23321.068823] printk: 475 messages 
suppressed.
Feb 15 12:23:36 venus kernel: [23321.068831] lost page write due to I/O 
error on dm-0
Feb 15 12:23:36 venus kernel: [23321.068842] lost page write due to I/O 
error on dm-0
Feb 15 12:23:38 venus kernel: [23323.251000] scsi 1:0:0:0: 
Direct-Access     WD       5000AAK External 1.06 PQ: 0 ANSI: 0
Feb 15 12:23:38 venus kernel: [23323.253219] sd 1:0:0:0: [sdb] 976773168 
512-byte hardware sectors (500108 MB)
Feb 15 12:23:38 venus kernel: [23323.253842] sd 1:0:0:0: [sdb] Write 
Protect is off
Feb 15 12:23:38 venus kernel: [23323.256121] sd 1:0:0:0: [sdb] 976773168 
512-byte hardware sectors (500108 MB)
Feb 15 12:23:38 venus kernel: [23323.256836] sd 1:0:0:0: [sdb] Write 
Protect is off
Feb 15 12:23:38 venus kernel: [23323.256849]  sdb: sdb1
Feb 15 12:23:38 venus kernel: [23323.258946] sd 1:0:0:0: [sdb] Attached 
SCSI disk
Feb 15 12:23:38 venus kernel: [23323.259003] sd 1:0:0:0: Attached scsi 
generic sg0 type 0

... and a few minutes later the wife (at home) patiently reboots the 
computer, and wonders why I got her hooked on this silly machine.

This issue is that the USB bus resets, and when the external disk is on 
an LVM2 volume, the dm entry is not removed.  When the USB disk comes 
back online, it gets another /dev/sd? entry, and cannot be accessed 
until we've done some cleanup, as in (as root, of course):

/etc/init.d/myth-backend stop
umount /storage
dmsetup delete <LVName>
vgscan
lvscan
vgchange -ay
mount /storage
/etc/init.d/myth-backend start

Phew!  Now ... I've considered adding this as a script to a 
/etc/udev/rules.d/10-local.rules file (to match the serial number of the 
disk, for example), but this is not a solution, only a way to let my 
wife only notice the issues minimally (though it would have the side 
effect of chopping up recordings into pieces with the backend restarts)..

Note that the previous Fedora setup had the storage partition on a 
primary partition (not LVM), so the USB drive would (upon reconnect) end 
up with the same letter, and reading/writing from myth-backend would 
simply resume after a pause.

Some other things I've tried:
boot the kernel with
acpi=off (no noticable difference)
Add the following to /etc/udev/rules.d/10-local.rules
SYSFS{serial}=="574341533830343230313137",RUN+="/bin/sh -c '/bin/echo 64 
 > /sys/block/%k/device/max_sectors'"

Some of these were suggested in 
(https://bugs.launchpad.net/ubuntu/+source/linux-source-2.6.20/+bug/61235)

Picking through
sdparm -al /dev/sda
shows '0' for IDLE and STANDBY entries.

Note that the problem appeared to be much more pronounced (resetting 
every 10's of minutes instead of one to three times per day) when I had 
the PCI bus underclocked in the BIOS ... I have since gotten a better 
cooling solution, and no longer need to do that for stability.

Finally, downgrading the usb to 1.1 (modprobe -r ehci_hcd) appeared to 
solve the USB resets, but there was not quite enough bandwidth left to 
both play and record to the disk.  (Playback would skip every couple of 
seconds ... the kids didn't care, but it'd drive my wife and I nuts!).

Any hints?

Thanks,

Josh