I looked into that by going over the results of 'grep bme /etc', so I was of course aware of that. Why do you think I suggested 'stop bme', and why I suggested to Titan (and others) to remove bme references in init.d/upstart-event scripts?
It's pretty clear bme needs to be started first, if it wants to tear down the device. So if we don't start it in init/upstart it probably won't reboot the kernel.

I'm a little unsure though about that (from top):
[2010-05-13 16:25:28] <DocScrutinizer> t-tan: [2010-05-13 00:41:14] <DocScrutinizer> PP:hald-runer PID:821 hald-addon-usb-cable: listening on /sys/devices/platform/musb_hdrc/usb1/../mode

I.E. we don't know what relies on a working BME and in which way.

There's one thing we can be sure though: battery will not explode or otherwise be killed when there's no bme. The bq24150 will wake up to sane safe defaults like max USB current 100mA, max battery charge voltage 3.7V, and it should start charging with these parameters, for at least 32min ("recover from flat battery" scheme). With a little luck (and support from a correctly configured 1707 PHY chip) the bq24150 will detect a dedicated charger (D+/- short) and crank up max USB current to 500mA. But no matter if it starts charging, and at which rate - it never will overcharge the battery unless there's a bme forcing it to do that. So removing / replacing bme is as safe as any other software project, you can do nasty things if you *want*, but they won't happen by accident. A failing or nonexistent bme will NOT cause any hazard.

/j