This explains some strangness I was seeing. From your description I understand why I can't probe across multiple sources, but just probing only the LED seems to trigger the system and I don't understand why. If I'm only using one probe it shouldn't matter if the scope sees ground as the point between the LED and mosfet, right? There is something fundamental about the scope's probe I still don't understand...
To a first degree -- if you have an isolated supply, or more basically, no other ground connections to the circuit -- yes, you can clip the probe to the switched node.
BUT, you'll still get a waveform similar to what's pictured, because there is no such thing as perfect (as in radio frequency) isolation. The PS, and scope (if ground is lifted), have capacitance to ground (or mains), so you get a long loop, from probe to scope cord to power supply cord to supply leads, which has a modest inductance (some uH). The capacitance and inductance resonate, and you get ringing, and currents induced in the circuit that wouldn't otherwise be there (your measurement disturbs the circuit -- no, it's not just for quantum mechanics!).
The preferred way is to keep grounds...groundly. The +30V and 0V nodes will have little AC on them, so are a good reference to work from. (One or the other, not both, of course!) The probe tip then can be clipped onto whatever circuit node is of interest.
Note that, if you ground to +30V, all signals are "below" ground, so you will see negative voltages. Just make the mental sign-flip (or tick the "invert" option on the input channel menu, if you have one?).
So that solves the 2nd order problem (when is isolation, not isolation? -- capacitance!).
You're still not completely free of ground noise, because there could be some crossed paths yet, at high frequency. Remember that grounding to the switching node. puts switching voltage on that huge loop? Well, any segment of that loop has a fraction of the full loop's inductance. That is to say, a loop is a single inductor, which is equivalent to many small inductors in series. Inductors in series means you have a voltage divider among them. So, even the few inches of probe ground clip, has a nonzero voltage drop! Or even the few cm of traces in the circuit, or the few mm of lead length on the components (and the bond wires inside them).
You probably won't have to worry about extremely low probing inductance now (the spring clip Mike's talking about) -- but that is definitely the preferred method.
By the way, the probe itself has some impedance, which draws current from the circuit, and that current returns through the probe ground (clip or spring, whichever you're using). Which, since that ground is a wire of nonzero length, it will also have some inductance, and some voltage drop at high frequencies. That voltage, in turn, still appears across the big loop (probe cable, mains cable, back to the supply), so you can still experience ringing from that path, even though your probing technique is otherwise quite good, and your ground point is otherwise quiet. In this case, you already have one intended current path: the path through the probe ground clip. You can reduce the current through other paths (the large loop) by increasing the impedance of those paths. How? Add ferrite beads to the probe cord.
Note this is a 3rd order consideration -- usually not significant, but it is nice to understand what's going on, especially when you're working with frequencies high enough (risetimes short enough) that you need a spring clip and whatnot.
(Basically, if you don't see a difference from clipping a ferrite bead on the probe cord, it's probably weak enough not to matter, and the 2nd order approximation will be fine.)
Tim