This is a theoretical question because I've never set up remote kernel debugging before -- but I will do at some point, which should hopefully answer some of the experimental questions I have.
What happens if you put a breakpoint in the breakpoint trap handling or kdcom / kdnet itself
I can't find a single thing about this, but in the former case, wouldn't the CPU just freeze because the breakpoint is continually being hit without the remote debugger getting chance to remove the breakpoint or iretting to the instruction after the breakpoint.
Also, from what I'm seeing, it seems like a stack trace hides any of the trap handling, and shows the breakpoint as the top frame on the stack.