In these last two dumps, it looks like indeed stdflt.sys is passing incorrect memory information to the stack, causing the next functions to fail. In both, the RDX register is supposed to be populated with the virtual address to be freed, but in both scenarios this register points to invalid VA. Again, in both the stack has just exited some routine in stdflt.sys, and given there's nothing else before it on the stack in either (other than base thread starts), I'm going to say it's a pretty obvious culprit here. Stdflt.sys' driver
timestamp is Thu Jul 16 11:36:50 2009, and from what I gather (and reading this thread), this is the accelerometer driver for the HDD.
It might be worth removing this package to remove the driver for a bit to see if it indeed makes the problem go away (it should).