This discussion is just a little nauseating for this site — 68kmla my ass. Thus, please excuse the following rant.
The truth is that the switch to Intel and associated developments in the land of software (Wirth's Law, Flash, and the relentless onward march—but to where?—of the web in general) have made G5 towers redundant in the popular imagination, which has the knock-on effect of making them cheap, plentiful and (in practical terms) perfectly collectable for the rest of us. To that extent, the state of the G5 presents a welcome opportunity.
The machines are, in a nutshell, vintage by computer standards (much of which is a stitch-up and a con, but I'll not go there just yet). Now, in the land of vintage computers, which we are all on here to talk about, I prefer to use professional-grade hardware when I can (Quadra 950s over LC475s, for example). Among more recent machines, the G5 towers fit that particular description, rather better than, say, might an early MacBook. A single-processor G5 maybe doesn't, admittedly, but a 2.3GHz duallie certainly does. Unless I had compelling issues related to portability, therefore, I'd take such a G5 tower over an early MacBook any day, essentially because the former will outlive the latter, has dollups more class, and in that sense and more kicks early MacBook butt. Speed is frankly over-rated; so is wasting one's youth on Flash and online video. I prefer interesting engineering and design quality.
There is a workable solution to the conundrum that arises when perfectly good tools are being displaced by the ephemeral lure of having and holding ever-newer, faster ones — don't buy new if you can avoid it. Be a Luddite, resist the hype, maximize the astonishingly untapped potential of the older tools (e.g., by learning to program in a TenFour Fox kind of way, or by buying abandoned software for them — like, say, Logic for ppc, or learn to compile for OSX on a risk-free box) and so reuse what others discard, before—eventually, admittedly—you yourself have to recycle it. Run the right software on it (which is what really matters), and make the computer industry poorer, while keeping yourself richer in the bargain. You will lose out on almost nothing, apart from having been had.
A perfectly functional professional-grade computer tower ought to keep getting used as long as possible, like any $3000+ tool in a rightly-ordered world. In what other walk of life do we throw such tools away so readily? It would be seen as barking madness in, say, a woodworking shop, or even in a hospital, and it ought also be be taken as madness in our homes and on our desks. Yes, there comes a point when technologies need to be replaced, but there also comes a point at which the recognition should dawn that Lemming-like behaviours are endemic in the world of computer consumption and marketing, and that they are best resisted. Yup,
that's us (you can still watch that in HTML5 on yer G5, BTW).
What to do with a G5 tower? Put it together with another one that has been abandoned, like I said, and make one better than both. Or in its current state, find someone else who can use it if you can't, even if it is only in order for them to experiment with OSX or Fedora (ported to PPC again) or with compiling, and thus give it away ("He who has two cloaks should give to him who has none"). That would really be something remarkable to do with a G5.
My $.02.