I'm for prong collars in most instances, be it a working dog or a training client. I'm also all for having good timing for balancing correction vs reward, and proper use.
I think its a common conception that reactive dogs are not candidates for prong collars because it increases reactivity due to "pain" of the correction, and I personally disagree for the majority of those cases.
As an example, I recently had a training client with a dog aggressive dog. Has a history of attacking other dogs. The woman is in her 50's and the dog was 15 month old, 80lb West German Showline GSD. The dog came to me with LAT training and a Halti from the previous trainer of 1 year, and was recommended to me for my experience working with the breed. Upon observation, this dog would offer attention at various distances for food reward, but the threshold under 10 feet was unattainable, and at this point, the dog had previously lunged against the head collar and was able to not only injure the handler in pursuit of another dog, but inflict multiple wounds on the dogs she made contact with.
Now sure, this woman could continually work on LAT training to break that threshold. We are now talking extensive months of training with an injured handler and a dog capable of pulling this woman to the ground. The dog would also have to be muzzled, in case she was able to pull to the other dog or break away from the handler (requiring additional home training of desensitizing to the muzzle).
This woman needs to be able to, at the very least, hold the dog safely on leash. Introduce the prong collar. Another dog is introduced at a distance within the dogs reaction threshold. As the dog approaches, she is corrected and removed from the situation for showing any excitable energy. One correction, walk away and done. If this is timed correctly and rewarded appropriately for the dog either redirecting attention or avoiding engaging in any response toward the other dog, then she is rewarded. This dog should go forward to show avoidance behavior toward the other dog. No "freaking out", no reaction, just avoidance because they have an understanding that their excited and reacting state of mind results in a correction. When they display this avoidance, the reward comes, and the dog learns that the avoidance of that behavior is what produces that reward. All at the same time, the dog will self correct for lunging and will be much less capable of knocking the owner to the ground.
Obviously there are dogs that are not going to respond to every training method. Plenty may respond with LAT and +reinforcement alone, but when its a large and dangerous dog, I don't try and baby foot around the issue. My main concern is applying the necessary control to keep that dog, the handler, and anything around them safe. And if the dog understands a black and white correction of what not to do, then its the safer alternative. And obviously any misuse and constant pressure/nagging corrections on a prong collar is not something that I'm an advocate of.
I also don't consider head collars "non-aversive". There are plenty of dogs that refuse to be conditioned to a head collar on a lead, and I've seen teeth knocked out, blown blood vessels in the eyes and nose, smashing heads against the ground etc in avoidance of a head collar.
I do use +reinforcement with a dog who has a higher level of drive to work for that reward. But if they don't want the cookie, I'm not shoving it down their throat to call it positive only training.