I've read that I need a bandwidth five times the CPU clock and ten times the CPU clock for pulses to look reasonably square.

Bandwidth should at least twice the base frequency. I'm not quite clear whether the CPU is really clocked with the base frequency directly - the 68060 will not accept such high frequencies, but Thomas should know better.

To see a real square wave (and not just the sine, as you would get with sampling at the Nyquist rate), you would require theoretically infinite bandwidth, but I believe you're reasonably off with five to ten times the frequency.

But there are probably more important issues, namely a reliable trigger capability (since signals are all digital) and having a memory scope would be considerably helpful as signals are usually not periodic, but single time (or rather, have a too long period to be useful on a regular analog scope).

At least, this would be my shopping list, maybe others could add.

Needless to say, that's probably beyond my budget, but you asked...

So long,
Thomas