Thanks Mike Most for saying what needed saying. While there is certainly still road to cover in getting everything perfectly in place for ACES to work as designed, this idea that it's hard and costly needs to die. There is nothing complicated in implementing it (designing is another story). I think that's evidenced by many of the people here talking about their path from camera to screen, many of which look an awful lot like brew-you-own version of ACES.
Half-float EXR is NOT necessary to derive most of the benefit of ACES. Using it is just icing on the cake. It really is a simple concept. Any image acquisition device -> IDT -> RRT/ODT. This is essentially how many, many people already work, just without standardized transforms. It'll get there, probably sooner than later. But many people are scared, I think partially because of all the talk about EXR containers. Fact is, you don't need them. Working in EXR would and will be great, but if the space and processing requirements are too much, don't sweat it. The core concepts of ACES work perfectly fine without it.