I think the basic problem with the fuse protecting the transistor is the low resistance of the circuit. So if the FET happens to turn all the way on like it did when Jeremy shorted the gate to +12 the current is only limited by the .13 ohm resistor and the output cap in his big power supply.
Going with this thought...
Rds-on for the FETs in question is 0.018 Ω, that would take 129 or so amps to reach 300 watts power dissipation in accordance with I[SUP]2[/SUP]R. The total circuit resistance with an FET fully on is 0.13 + 0.018 Ohms, or ~148 milliohms. Thus the max current is 24 Volts divided by 0.148, or ~162 Amps. Which is enough over the limit to be a problem.
We could add in the stray resistance of the wiring, but that isn't a good idea as it is probably somewhat inductive. If we use smaller value current sense resistors, the problem will be even worse.
Again, a full crowbar would probably allow us to side step the problem because it would let us sense over current when it happens, and then try to share it across all the FETs temporarily. This system wide over current could be made to pop a single large main fuse, or even better, a resetable breaker. On the flip side of the coin, a crowbar circuit is more parts added and more designing that we would have to do. It also has the unsavory effect of short circuiting the system for a small time. So I fully understand any reluctance from the group to go this route.
I don't hate the idea of using the resistors as the fuse either mind you. It's nice to make parts dual propose, and if they were intended to do it then even better. The resistor is where all the power will go if a FET shorts anyway, that should virtually guarantee that it will blow. The only problem with this situation is if just one FET goes belly up, it may be a while before jocanon notices the discrepancy, as the system as a whole should still limp along. This could lead to improperly tested PSUs, or the rest of the dummy load bearing a larger part of the work than we have designed. Worst case scenario the old possibility of a cascading failure comes back.