What I ended up doing was to use a seperate sense resistor and op-amp for each mosfet. That way, each mosfets share of the total is controled by it's own closed loop.
Yes, it's trivial to do that too. So that's what we are probably going to run with after all.
8 FET's, 8 Sense resistors, 2 LM324 or similar, 1 pot and other resistor for the load adjustment, and we are set to build a dummy load. You can also get 10 of each of the FET's and sense resistors and 5 dual Op-Amps if you want to get the price drops for buying more than 10 parts. It will be marginally more money total, but it will be a better value.
Speaking of value, Watt for Watt, BJT's are about 1.5X more expensive than MOSFET's are. So it would still be more than likely better to use FET's, even if such a circuit requires marginally more parts. The construction will still be mostly the same no mater what you do.
I'm may not attempt to try the liner way after all. The more and more I look at it, plus all the commentary makes me convinced it will blow easy. So it's down to these three circuits then...
1) MOSFET load with 1 Op-Amp and sense resistor per FET.
2) BJT's, with one resistor per unit, but one amp circuit for the base current. (Yet to be discussed)
3) Resistor bank, with high current moving linkages, or switches. (you can use MOSFET's for the switches BTW)
The other options that are lost, and why are...
1) PWM fixed resistive load.
2) Parallel linear MOSFET's with one Op-Amp.
The PWM method will function. It may even function without the filter, I can't be for sure. But if it does need the filtering it will be expensive. This method is intrinsicaly safe though, if for what ever reason the FET's go full short, it will be caught by the resistors, that will only allow a set in stone amount of current to pass. This circuit is still it is somewhat complex to build too.
The linear MOSFET bank would work *IF* you sufficiently overrated the parts so as to make any part tolerance differences be covered by the slack in Wattage. Also, a working PSU *MIGHT* catch any failures, do to over current protection. (MOSFET's usually fail short circuit) However, to make the best circuit possible, it makes sense to make it useful for more than an over current protected supply. And it is always most efficient money wise to use the most Wattage you can safely get out of power parts. Other methods will allow safer operations at higher levels of power for around the same price.
Anyway, I'm most likely going to change my experiment to test the independent control loop based MOSFET load, as laid out previously. I may still try the first MOSFET method if I feel like probably destroying some MOSFET's.