I don't doubt on the potential of a FCPx on the long run. But the problem is that we are in between 2 worlds, 2 budgets and 2 generations; where the jobs are it's the old way wich is basically a cascade of roundtripping; and roundtripping is very much metadatas-asset management based workflow orientated on specialities. Right now we are in the wild west.
We know that this kind of workflow is going to change inevitably, time+tech always simplify and reduce costs, as well as the fragmentated pipeline. Soon, cutting in Avid, color in Resolve, compo in Nuke, conform and finish in Scratch, tricking metadatas in text editors will be obsolete. But this "soon" could be in 2 years or in 15years.
The only brand I can think of that seems to have a cutting edge view on the future is Adobe in the sense that the big shift that is happening is that we are asked to be more and more multitask. As software and HW engineering evolve, they allow for a single person or a very very reduced team to cover the all pipeline, even in the demanding high-end. What were let's say 10 seats in a prod house will be tomorrow one or 2. Will FCPx cover the all pipeline aspect (?), it will be a winner IMO, but if it's "only" a wisedly implemented NLE, kids with carreer view in motion will more likely to follow Adobe because the days of editing-only are numbered, unless...
I remember for example a Michael Reichmann's Key made with FCPx. It was really impressive for such a cheap software. Nothing or very little to envie to a spectramatte. And it seemed straightforward, and straightforward is good but can also be a trap if it lacks controls. The question is: how does it works with conflictive plates? how does it works when things get tough? when man has to take control of the machine? Do I have to naughtyroundtripping again to a Nuke or an AE to save the furnitures from the fire? and then how stable the roundtrip is? (as an example, the keyer in Avid is enough powerfull so it was used all the time within the edit of the Marvels) Is FCPx suitable for handeling securely a huge amount of datas? that's another question, and probably one of the most important in fact if the target is professional etc etc...
So IMO, it will depend very much on how they make evolve the software in the next years. The brain seems brilliant, but not yet able to cover-convinced the industry needs from bottom to top with enough reliability, and maybe also because there is a big amount of resistence to change habits. With Adobe, it's already there to some extend, it smells quite strong that the suite is going to be a serious option for covering all the pipeline, included a strong compositing app. But I do recognize that Adobe has been developped with the "old way" in mind, the strengh resides in the simplified roundtrip (but not lack of) and the interface being of the same house. IMO, the PP editor sucks in front of an Avid or FCP, but if you take the all suite together then that's another story and way cheaper than a DS, it's easy to forgive the editor's design. If FCPx revolutioned the "editing way" but lacks in stability, in asset management etc etc...it will be difficult to be accepted by the coming or next gen houses. Time will tell. But I understand you perfectly why you are intrigued by it.
The big questions as you pointed are the MacPros roadmap. Everybody is asking this question. Before asking it in the forum, I talked to some Mac users here and same as you, they could not really answer because they were asking themselves similar questions. Some years ago, I would have asked this question into the Mac users crowd and they would have answered inmediatly: switch back to Mac with confidence. Now I don't see this confidence anymore among the editors Mac users and that's why it makes me wonder.
As you pointed, not an easy decision because there are no easy answers.
Ps: as you talked about FCPx, it reminds me the GH2 saga. The GH2 was way more powerfull than the 5D2 for motion, but despite that it has never really been embrassed by the industry and not just because of the crop factor (the AF100 is well respected) but because of a series of bad implementations and some house issues and in the end the 5D2, although being less refined for video was more suitable in the field.