Back in the days of tube audio amplifiers, we tested them for output power by replacing the speaker with a resistive load
This load was made up of a higher value resistor & a lower value one "in series".
In use, we measured across the lower valued resistor, to minimise any loading upon the value of the overall resistance of the load.
We usually used an analog VOM to read the RMS voltage to determine the power output, which had a much lower input impedance than things like DMMs, VTVMs & Oscilloscopes.
Because we were reading a fraction of the voltage across the total load, we had to use a "correction factor".
This was easily determined, as we knew the ratio of the lower resistance to the overall resistance (plain old voltage divider theory).
With the calculated total voltage, we used "V^2/R" to calculate the power output.
We did hang an oscilloscope across the same points on the load, to determine the clipping point.
Another thing we did was, using a Noise & Distortion meter bridged across the same points measure those parameters at more normal operation levels.
The VOM, like a DMM, didn't have one lead grounded, neither did the N & D set, but a 'scope or VTVM did.
Tube amplifiers used output transformers, the speaker terminals were not connected to ground, so there was no problem.
Even with the early solid state amps, where the output had one side grounded, as long as you checked that, you were "good to go".
Amplifiers with bridged output stages are another story---- connect either speaker output terminal to ground, & you can damage your amplifier.
There are two workarounds:-
(1)The "proper" way--------buy a differential probe.
These are not as horrendously expensive as they once were, but the cost is not inconsiderable, either.
(2) Using two channels of your 'scope, hang the probes off both ends of your load resistor, (with both ground clips to a convenient ground) & using, for instance, the "channel 1 plus channel 2" setting of your 'scope, you can see the voltage across the load resistor.
This is commonly known as the "pseudo-differential" connection, & is usually frowned on by the purists in this forum , but for tests of the level you are attempting, will give satisfactory results.