The second mysterious failure is really annoying, since it's in the area that I was just asked to fix. I am told that it is not a problem, because I can always make a release build and install it on a temporary test machine to see how it is working -- and if it is not working, I suppose I can use 1970s debugging techniques to figure out why it isn't working.
Thus do we turn a debugging process that should have rapid turnaround into one that, well, does not.
I have complained about this. :)