Predicted for over a decade now, has never come to pass, nor is it likely to. My college CS department was heavily Linux flavored, and I still think Linux is a joke as an end user OS, no matter what kind of GUI sugar you put on top of it. It's just too technically oriented, and the so-called Open Source community is too arrogant and insular to ever invest the time and resources that either Apple or MS has to create a GUI that covers all of the bases that the average user needs covered.
I generally agree with your assessment of Linux on the desktop for general users, though of course its going to stay strong in the developer and science/engineering communities. But there seems to be a perceptible sea change in the Linux community, or perhaps a fork. There is a growing camp of enthusiasts bent on making Linux easier, in opposition to the elitist crowd. Its certainly not inevitable, but I think its entirely possible that the user friendly Linux crowd will get to the point where they just have to replace X, and after the civil war is over and the horses burned I think there will be real progress made. Now whether they will catch up to MS and Apple, who knows. My feeling is that a lot of professed faithful, when faced with the schism, will find themselves siding with the New Testament apostles. Can't think of any more metaphors to throw in the stew.
And if the GUI wrapping Linux just provides basic apps and a good browser, then webbed apps can provide all the needed functionality without requiring high school dropouts to wrestle with Excel macros.
I will agree that a lot of databased applications will be delivered by web browser technology in the future -- there are just too many good business reasons not to. But that is not the same thing as saying that those browsers will be running in a Linux environment. Standard productivity apps require too much interactivity to ever be viable over the network, no matter what some guru may have whispered in your ear about application servers. And as long as productivity remains local, the OS is going to have to be keyed to the average hourly clerical employee, not to what some IT geek thinks is the perfect Linux setup for accessing the company intranet.
My theory, though, is that the majority of users in the work environment need only extremely limited productivity apps (i.e. MS-Office). Sure, white collar people doing real IP work need those things, but the vast majority of businesses have people who just use PCs for e-mail, diddling with the accounting/inventory program, calendaring, and POS operations. The average salesdroid, paper pusher, or line supervisor doesn't need animated slide transitions or any of that crap. I figure the Office user of 2015 is going to be the archetypical power user, maybe 20% or less of the total desktop population. Rolling out Office to the bottom half of the bell curve is not going to improve productivity I don't think.
Now, I don't believe the distant application server concept is particularly good either. Every hiccup in DNS service could dump the session into the bit bucket. The rational solution, methinks, is a return to the minicomputer concept, more or less. Mostly thin clients connected to a departmental server, with loaded PCs given to those who really need them. Seems to me that admin can still be contracted out, dunno about Win2k3 but Linux/Unix one can do most of what needs doing thru the firewall using slogin (yes, I realize tunneled X sessions suck, but Real Men use the CLI anyway...).