This is a 260.xx-specific bug, where the driver crashes under some very specific circumstancse:
1. you create two threads
2. you load OpenGL dynamically on thread 1 (instead of linking with -lGL)
3. you create an OpenGL context on each thread
If all three hold, the crash occurs as soon as you call MakeCurrent on either thread. If you load OpenGL before creating the 2nd thread, the crash doesn't occur.
The issue is known to Nvidia, we are just waiting for a fix (ever since 260.99 was released).
Multithreading.cs. The initialization logic is here: WinGLContext.cs.
Toggling "threaded optimization" in nvidia settings avoids this issue. This thread [opengl.org] and this thread [nvnews] also refer to this issue (there are more).
Unfortunately, I cannot install beta drivers on my workstation. I'll try to build a scratch machine for testing.Have you tried the 265.90 drivers?