IF the baud rates are precisely matched at both ends, to within the usual commercial tolerance on crystal oscillators, i.e, 100 ppm, then the effects of crosstalk, induced noise, etc, on data tranception are negligible so long as the combination of baud rates and cable lengths are kept within specified limits.

Shot noise and other single-event-upsets do cause errors, and there does, as you suggest, need to be some protocol for handling them. Generally, parity is adequate to detect them.

The matching of baud rates, when there are ten or a dozen bits involved, is not a critical factor, since the timing starts over at each start bit. For 10 bits, 5% error between stations should be barely acceptable. 4.5% should work 100% of the time.

RE