ROM files from OP and MPU messages from the above log will boot the camera in QEMU, although with main display off, but menu works on the virtual camera. So, I'm not sure what's going on.
The above error is actually a bug in my logging code, or - rather - a side effect of the overhead of capturing debug messages. Small timing differences in PropMgr task could cause this message on this particular camera, at least.
Something unusual: log files are saved if and only if an ERR70 appears on the main screen. Otherwise, the camera probably does not initialize file I/O for some reason. To be investigated, but I cannot exclude the power button at this stage.
I've tried to compare the MPU messages with the power button in both positions (on or off), from my 60D, but couldn't find any difference. Rather, it seems to me that MPU decides to power down the camera while main CPU (ICU) code is still executing (early boot), but before the SD card is initialized. There may be some event or flag or whatever, that allows the MPU to power down the camera, but I haven't identified it yet.