Probably the extra capacitance delays the signal so the latch internal to the programmer gets a valid data level, most likely it is being latched just as the signal is being changed, and depending on the device, temperature and supply voltage it will read one or the other. The delay slows it down to a time afterwards when it is stable,
Had plenty of fun with this on cards that would nominally be perfectly functional, but just would not work with each other. Change one or the other and it works. Then take to operating temp of 120C and see if it still works, if not try another card. then take to the ATE and let it run through diagnostics for 5 hours and see if it fails, generally not, but it might do so again in a year.