It's actually relatively common knowledge that Jobs himself would have preferred to stick with PowerPC hardware. Steve Jobs was, for all the growing he'd done at NeXT, still fairly capricious and was heavily susceptible to the belief that anything "not invented here" was a bad idea.
The problem is that in the real world, nobody had a good POWER or PowerPC road-map for what Apple needed to do, which was portable computers on one end of things, and computers that were "actually fast" on the other end of things. In 2005, Apple would still have found itself "competing" with $10,000 RISC UNIX workstations (the ones that would all die off in the next year or so, but hadn't quite yet), and so part of keeping a product like the Power Macintosh G5 relevant would have involved not slotting it in with the machines in the "five digit price tag" category. (Think: anything worth buying from Sun, HP(DEC/Compaq), IBM, SGI et al.) Unfortunately, part of doing that really meant using parts that just weren't actually as fast and good as what IBM was using in their own systems.
In terms of Mac OS X requirements -- (2000-2005) Part of the problem is that PowerPC was evolving extremely rapidly at that point. It had to in order ot keep up with Intel. Pentium 4s, for as terrible as they were, often massively out-performed G4s, and a good late-gen Pentium 4 on a 900-series chipset with DDR2 RAM was a force to be reckoned with, in terms of raw performance, and even moreso, in terms of performance per dollar.
With the rise of the web, and with information about Intel's road-map in the public eye, I'd like to think Apple had the foresight to know that PowerPC wasn't going to get them what they needed. this wasn't the '80s any more and you can't sell the same CPU from 1984 to 1990 and expect people not to look at you funny.
And, Apple was desperate. They weren't quite at the point of releasing a different-looking Power Macintosh with literally identical guts as the last generation (complete with an introduction video!) but I think Apple knew they were behind. As the 2000s wore on, the web started to be a thing that really drove performance improvements. Without that, Macs could probably have plodded on as productivity and media machines, with the default narrative being that you buy a Mac not because it's fast, but because it gets your work done well or because they're reliable.
Although the exact cut-offs were often arbitrary, I think that they happened because Apple needed a convenient way for people to self-identify whether or not their machine would run a program or an OS. An oft-cited example is Mac OS X 10.3 or 10.4, which "required" a firewire port. It did not actually need a firewire port to run, but the firewire port designation was convenient because most G3 Macs that didn't have firewire also lacked a powerful enough GPU, a big enough disk, and enough RAM to run the software well. And so, one or two machine that probably could have done well got cut off with a whole group of them that just had no chance.
In terms of dumping PowerPC Macs once the Intel switch was made? They got a few years. It wasn't until Mac OS X 10.7.0 was released that the last of the PPC Macs stopped getting software patches from Apple, and I think most developers had already given up of their own accord by then, not at Apple's behest. Microsoft, for example, has long had a habit of making Office available only for the most recent one or two versions of Mac OS X. Adobe has been the same, and I think part of that has been that classically, Office and Creative Suite had very long life-cycles. Adobe's release cycle quickened a lot during the 2000s and early 2010s but Microsoft's doesn't really show that much indication of speeding up.
G5s may not have "a million" cores, but mainstream PCs don't either. The only things with more than four cores at this point are enthusiast platforms, workstations, and mobile phone platforms that have "eight cores" (whereby it's two quad-core designs glued together, one designed for performance, and one designed for low power usage.)
Ultimately, the problem with calling the G5 a really useful computer is that for $700 you can pick up something with a 5-watt dual core processor running at less than a gigahert (base frequency) that will happily trounce the G5 at everything. You might need to tdpUp the chip or move to a 15-watt chip to actually be twice as fast as a quad, but it's not going to be hard or expensive.
The G5 served a purpose in the 2000s. So did the Octane2 and so did the C8000 and so did the Ultra45. Suggesting that it's a suitable computer for a casual computer user today, however, just isn't in the cards. "Barely able to keep up with modern netbooks," is probably a better way to talk about G5 performance (let alone G4 performance) and the most relevance it has is either doing whatever it did in 2005, or working for a hobbyist.
At the end of the day, it's somewhat disingenuous to talk about the very last generation of Power Macintosh G5s as though they're at all representative of anything else Apple had built up to that point. Physical reliability aside, the Quad is over twice as fast as every other G5, including the dual 2.7GHz model. The other models are basically in line with normal G5 performance, close to the top of the heap, but they didn't exactly set the world on fire. Normal G5 performance is pretty close to what you'd see out of that 2.0GHz dual-core G5. Even more systems were dual 1.8GHz models. And so, thinking about comparisons to that system are probably better.
So suggesting that just because the Quad (if you equip it with the best of everything) is well suited to a particular task doesn't mean any other G5 (especially those that can't run 16 gigs of RAM) are suited to that task. They might be able to do it, but it won't necessarily be fun.