I have an interesting observation to make.   Given the problem was related to a specific instance of the vmdk on disk, and chkdsk showed no errors, I thought I would look at the fragmentation state of the host disk.   Indeed the host disk was fragmented.   So, first I again validated the problem existed with one particular instance of the vmdk on disk.   With that instance the exception occured 100% of the time on many attempts (> 50).   The vmdk itself is 22 GB in size.  The failure occurrs at different specific times around ten minutes into the startup sequence of nine Windows services in the guest.   Then I defragged the host disk overnight.   Then I started the VM with that same vmdk and the problem does not occur at all any more (so far more than >10 attempts but now I am getting tired of running it).

Is it possible there is a problem in VMware 6.5 related to a large vmdk stored on a fragmented disk?

Other explanations are possible (like a timing related bug in one of the Windows services I am starting that happens to be affected by the defrag). 

But I would be interested in hearing if anyone else has encountered issues related to fragmentation.