Reliance
When I cast my mind back to the late 80's and early 90's, I recall, during my time as a computer support engineer, the vast diversity of operating system "platforms" that were in use in my home town.
I recall companies running DOS (maybe with Windows 3), GEM (on Atari STs), CP/M (on Apricots or XT based PC's) and CP/M on Amstrad PCWs. If you were the local ProntaPrint you ran Apple MACs.
As you got to know each company, you got to know what systems they were running. The one software application that they were *all* running appeared to Sage accounts and Sage Payroll.
I think the main reason for convergence to one O/S (this reason does not apply to the hardware platform) is the dominance of the internet. Before the web, companies ran as small independant outfits, often relying on other businesses. For example, when you needed paper for your photo-copiers, or ink, or floppy disks, or hardware, it was purchased locally.
Today, different companies exchange masses of information, daily, continually, in the form of email, written documents, spreadsheets etc. So as business has naturally diverged in terms of how and with whom they trade, they have converged in terms of information exchange. This is Darwinian 'natural selection' at it's best.
I personally would love to see a completely new operating system, not based on Windows or Linux. Sure, there would be no apps for it, but the 'break away' from our current (legacy?) based OS would be a breath of fresh air.
PSION had the chance, in my opinion. Their laptop/netbook OS was so good, it would have been a natural step to produce a desktop version of the OS. Imagine being able to purchase an application for your Psion netbook, and run the same installable on your desktop, with all the advantages of full syncrhonisation etc.
Beos could have been another contender, yet it fell by the wayside.
I'm not sure what the 'magic ingredient' for a new (sucessfull) OS would be though: Everyone that I have spoken to that has used BeOs has said how excellent it was, so technical ability is not the issue. I guess that only leaves the applications?
Thus, I'm forced to conclude that we won't see any major new OS for the desktop intel platform in recent years. To change would be akin to taking every car in the country and converting it to run on rails, instead of roads.
Sigh.