Again with this "it's overkill" and yet "it's not enough". What is it you expect your mythical developer to be doing that they need a terabyte of local, internal storage?
Well, as a developer myself I can explain:
My projects directory alone is about 500 GB. Project files, source code, image and video data (for ML training, etc). A terabyte is a reasonable amount of storage; 256 isn't. At that price its ridiculous to even argue about it. At 6000 starting price it should be included, period.
However, a Xeon including the logicboard IS overkill, for it can do nothing an i7 couldn't do from a developers point of view. ECC does not help either.
I never actually came across someone who benefitted from ECC RAM tbh. ECC may be a requirement for server applications, maybe rendering farms, I don't know, not my field of expertise. However, for development it just is not needed. It is very very likely safe to say that if an app crashes, or a compiler, or whatever, the cause was not the RAM being non-ECC.
So yes, the Mac Pro is overkill with respect to cpu and logicboard used and its at the same time inadequate in terms of storage and, potentially, depending on field, GPU. Or in other words, you could build a developer machine at roughly half the price without making any sacrifices whatsoever.
Its kind of strange to even argue about; assuming every developer and/or creative had to spend 3000 to throw out of the window we'd be living in an ideal world.
Unfortunately, most of us have to consider their spending. As a developer machine this is just very bad value for money. Not a bad machine by itself, of course, but the value preposition is not good at all.
Just compare the classic Mac Pro with the new one: in 2010 starting price was somewhere around 2500. Today its more than twice that, without providing that much more value. Yes, the CPU is fast. Just as it's been back then (relatively).
I'm not saying all developers can or will or should use, or want to use this. I'm saying this obsession with claiming that (a) it's too expensive for developers and (b) it doesn't have what developers use/need is a fallacy.
As I just explained, you are wrong here. Sure, some of the shortcomings can be overcome by spending even more money (on SSD, GPU, for example).
But even then, the machine cannot be used by developers in the machine learning field because this field is factually close to 100 % depending on CUDA.
Now, of course, this is not due a tech limitation of the Mac Pro itself. But for a developer or researcher (or college) this is irrelevant; she simply cannot use this machine for serious ML applications, at any price.