I saw an interesting thing said today (not from the "chat with our friends"), which is that the overwhelming feedback Apple received from those Mac Pro customers they have talked to in trying to get their heads around this, is that customers want Nvidia GPUs, and they want the ability to use multiple, full-fat off-the-shelf versions. That has been an unambiguous message - people don't want buying Apple pro hardware to be a risk that they'll be out of sync with the rest of the industry when the next disruption like the GTX1080 arrives. It doesn't matter how exciting Metal is, if Resolve is better with CUDA, then Apple needs a machine that can do CUDA, because if you use Resolve, your competition will buy that CUDA machine.
Pro customers don't want Apple to be "exciting" and "innovative" with tools they rely on to earn a living. Predictable and reliable seems to be a theme, and it was interesting to see Phil acknowledge that people who rely on Apple tools are hurt by Apple's screwup.
Apple were given an important lesson here - they can't control where the creative industries go, and when they get too far from the mainstream of how systems are architected, they can't make something better enough, that it's worth the risk if they get it wrong, and everyone else zags. They thought multiple low power GPUs would be the future, whereas it's actually been single, and multiple high-power setups.
They poured a boatload of money into developing the 2013 machine, and the bet didn't pay off - you'd have to hope the lesson learned is that they need to reduce their exposure to getting it wrong, and concentrate their resources on building the best scaffold, for hanging industry standard parts off.
It's a bit like the whole PPC / Intel transition, and now the arguments for AMD macs - it doesn't matter what argument is made about which is faster in a bakeoff or on paper - everything you do different to the competition eats marketing resources - if you've got the same CPU as everyone else, all you have to market is the things you do better.