kanoi, why does his work?
My guess is the problem depends on different 32bit implementations on the 64bit windows versions and the bug that I guess might be in different Windows 64bit->32bit Serial/USB code.
e.g. in Win7-64bit there is a limit of 8 USB devices when running 32 bit (XP) mode
(according to http://answers.microsoft.com/en-us/windows/forum/windows_7-windows_programs/windows-7-xp-mode-usb-limit/ff50b790-6e6e-47e4-8f48-5727896f5aeb
xDGDZEx had the problem when he hit 7 or 8 BFL's (not 9)https://bitcointalk.org/index.php?topic=80852.0
The first problem there actually suggests it may even be opening and closing valid devices multiple times (but it also happens on invalid devices)
The only change relevant to FPGA in the versions he compared was that Icarus opened the devices first (and failed) then BFL opened the devices but ran out before it got to 8 devices.
The previous version worked where it did the BFL code first so the BFL's all opened successfully
I came to the uncertain conclusion that the problem was that the open code in windows seems to sometimes identify bitforce:\\.\COMn as a USB name and allocated some resource to it but then failed to open it (as expected) but returned an error without always releasing the resource
I was hoping for xDGDZEx to test the change but I've not seen him in IRC since I last PM'd him.
The code change was to not attempt to open the invalid device names (and hopefully solve the problem if it is a windows bug)
That same code change worked for you bitpop so again I'm guessing it's related to a USB count issue in your different version of windows.
Either way, the change isn't a problem in any way (it simply stops doing the "expected to fail" opens) and if that solves the problems for some people - good