I'm probably going out on a limb here and slightly disagreeing, ok maybe not disagreeing but raising a few points...

In theory the secondary 0v rails can be joined to chassis ground/supply earth quite happily - no current can flow between this link *unless* there is a path from the +ve supply rail back to the chassis ground somewhere - this should never happen.
Now, i say "in theory" because the chassis ground may possibly contain all sorts of nasty things like RF from a plasma supply, spikes from other sources and generally some noise. The level of this electrical muck can depend on how good or bad your supply earth is - if it is good then the muck gets drained rapidly to safety, if bad then it can raise the earth potential above where you think it should be for the duration.
Next - suppose I have my low voltage supplies floating - this is what you describe - "floating" because neither rail is grounded, and there is a short from +ve to chassis - what will happen? Nothing will happen apart from all the muck on the earth will now get injected into the +ve supply rail and forwards.
So, say i ground my 0v rail to chassis and there is a +ve short to chassis - what will happen now? It *should* blow the fuse alerting me to an issue, after tearing my hair out for a few hours or days trying to find out why it stopped

As you can see, grounding can be a nightmare and can cause issues if not done correctly, I am not saying that 0v should be grounded, especially when a PC is involved UNLESS there is 100% opto-isolation on all signals and the 5v rail is not being pulled outside the PC to power a BOB or the MP3 player etc

Its all about loops, interference, safety, shielding etc and sometimes - take a CNC plasma cutter, probably the worst case scenario, it can be a nightmare to get right.
My Bridgeport CNC conversion has multiple PSU's - one is grounded and powers the control signals into and out of the CSMIO controller, it needs to be grounded because i use a touch-plate which uses ground through the machine as a return path.
The other PSU is NOT grounded and only powers the CSMIO controller itself, this controller has fully isolated inputs and outputs so this lack of grounding ensures there no possible path for any muck to get into the controller. The PC will be grounded of course but it uses ethernet which is isolated so no loop there. The servo drives are grounded via the chassis (bolted down) and the AC supplies but internally who knows?
