Ever since 3.99.7/8 my EHCI and BGE controllers have been acting up  
more than usual.

I've stated the clues to EHCI before, everything is normal in the  
dmesg except
ehci_sync_hc: tsleep() = 35

If 35 is errno, the man page says something about it being a  
temporary error condition, try again later. That doesn't make any  
sense to me but I don't really have a handle on the code either.

bge is doing:

bge0: watchdog timeout -- resetting
bge0: block failed to stop: reg 0x1400, bit 0x00000002
...

Its quite curious that they both seem to not work at the same time.  
Is this a symptom of not being able to get contiguous kernel memory  
for buffers? Or am I just "lucky"? The fact that this problem  
persists across multiple warm and cold boots and then suddenly works  
fine is also disturbing. Not being able to type or ssh in makes this  
box pretty much useless. Can someone please help me diagnose this?