My final take on this particular issue of the rising clock: Since I removed pull-up resistor R15 from the development board, I have not had any further issues with the flash programming. I put a comment in the git issue tracker to fix the code so that the pass-through implementation is robust in avoiding this issue by releasing control of the clock after releasing control of the chip select. But the main problem is that there never should have been a pull-up resistor for that signal on the board.