Ain't gonna get PTCs big enough for that directly, but you could put one in front of the transistor, or use an NTC to shunt its gate drive.
Preferably with a comparator, so the transistor doesn't spend extra time in the linear range, burning a sizable fraction of that 800W load. And with the comparator output feeding back to the controller to command it to stop braking, or warn the user, or some kind of useful feedback or control.
Also, might it be a... rather horrific idea, to consider disabling a braking element? I have no idea what this is braking, but if it's braking motive equipment as the name suggests, it's probably a good idea it comes to a safe, zero-energy condition under such a failure mode. In which case, the resistor should be designed to handle this easily in the first place, but also could be designed with a redundant backup perhaps, or using especially robust components (e.g., vitreous enamel or bare wirewound types versus metal-case types) with the understanding that, they will probably sustain one or a few excessive operations, but should be replaced promptly after such stress.
A good mission-critical example being commercial jet brakes. They have to handle the enormous energy of stopping, well, however many hundreds of ton(ne)s is moving at ~ludicrous ground speed. They can do it, magnificently so, but they get unbelievably hot in the process, just barely holding together as the craft comes to a halt. Brake fires are an immediate grounding order and maintenance step (AFAIK; or at least, I would hope so..?!). Normal landings of course (with long runways, and often, thrust reversers to help out) have enough time to keep them cool during use so only need regularly scheduled maintenance.
And if it's, like, an electric vehicle application, why not use a regenerative motor controller? Batteries make fantastic brake dumps, dissipating a fraction of the heat! The cheapest watt to dissipate is the watt you didn't have to dissipate in the first place.
Tim