The clamping diodes are a compromise - too high a current rating means a large junction area and too much capacitance for fast data signals. Therefore, worst case, if the user connects one of the data lines to +14V from the bench PSU, the clamping diode may not survive, but as long as the energy the polyfuse lets through isn't enough to blow the failed diode open, there should be no damage other than to the diode. I'd start by looking at BAT54S series Schottky diode pairs, which can withstand 600mA non repetitive peak If for no more than 10ms and see if that's enough for a reliable trip of a 50mA polyfuse. If not, that should at least give you some idea how beefy a diode you actually need.
The line drivers should be powered from the switched supply at the drain of the P-MOSFET, and that supply should be protected against overvoltage by a crowbar circuit with a 5.5V trip point that turns off the P-MOSFET and grounds the switched supply via a beefy N-MOSFET, leaving the polyfuses in the switched Vcc lines and the data lines to hopefully isolate them before damage occurs to the programmer or the target.
If your programmer had a higher voltage power source than the 5V USB Vbus supply, you'd have better options as the switched Vcc supply could come from a LDO with a current limit, an external Adj pin and an enable pin, eliminating the need for a P-MOSFET and permitting target power to be current limited at source, and allowing a Schottky diode to protect it against being back-fed from a powered target without excessive voltage drop. You'd still need the crowbar circuit and polyfuses to provide protection against misconnection to a powered target.