Not all scopes are plastic toy scopes . Older and high-end scopes tend to have a metal chassis bonded to earth. The OP's scope may very well be a big honking Tek 465 with nice big metal knobs. And even in the case of plastic scopes, do you think any testing went into isolation between the user and ground? Any metal surface near the user will be connected to ground (eg. LCD bezel). Those rotary encoders are not CAT II+ 400 V (vaporizing a ground lead does not sound like CAT I to me) rated. Especially with those dodgy knobs that were used on some Rigol scopes. Why do you think all manufacturers recommend against this procedure for high-energy circuits?
Yes, indeed. It's just when I read "scope" in this forum I automatically think "plastic DSO" such as Rigol DS1052E (which I own, by the way)
But at ordinary mains-socket (<=240V @ 20A) voltage and power levels, mostly any piece of plastic however thin will provide enough isolation, even an eventual crack on the knob will probably not matter as long you do not touch exposed metal directly.
Of course you shouldn't be careless, but safety is about making it hard to make fatal mistakes. Balancing on a cord 10 meters in the air can also be safe if you're careful, but that doesn't make it a good workplace.
If those kinds of energies are involved, then you shouldn't be touching the DUT or probe will powered. You should also use face protection and probably get the HV diff probe out.
I didn't say it would be safe, what I meant is that in the specific situation of probing a live (relative to earth) DUT, having the scope chassis earthed will not improve safety, and might (in the event of a short) even be worse, regardless of the energy levels involved (I admit there were some exaggeration of the consequences, and as a CAT I device, an ordinary DSO should not be used to probe mains-connected circuits anyway).
Another issue I don't get about floating your scope is that you lose all those advantages of a floating ground as soon as you clip on your first probe. Clipping the second probe to a different potential can still vaporize the ground lead and damage your DUT. If you're only using one scope anyhow, which not put the two channels in subtract mode for pseudo-differential operation?
Yes, the pseudo-differential operation is probably a (slightly) better idea, since using a second channel negates purpose of lifting the scope's ground.
Of course, all of what I said should be taken on a "hacking" perspective, based on self experience, and I don't mean any safety standard should be disregarded.