Most things that put off enough voltage to require one of those probes arc through air at that distance. If it's close enough for a clusmy person to short something, it's already either blown up or the insulation is thick enough to withstand some awkward poking with a probe (or its submerged in an oil or other nonpolar solvent).
So honestly, this demonstrates the value of an education in the field. The EE BS degree has a tremendous focus on theory and students don't really get a great appreciation for the reason that it has such a focus. But if you thought about it, you'd realize that if you're measuring, say, the voltage across a capacitor, there should never be any fast dv/dt spikes. The physics doesn't allow it, which means you're measuring something else. You can either hand wave and say how it can't be, and ignore the spikes, or you can set up the measurement properly to exclude them.
This later approach takes experience, which is harder to teach. Most of us pick it up from years in the lab, or occasionally reading an article from a master like Bob Pease.
It's a shame we haven't figured out a good blend between theory and practice that can produce a graduate who actually knows this kind of stuff. But I suppose the field of electrical engineering is so broad that if we tried, school would either take 10 years or would produce such a finely tuned graduate they'd be useless for anything else.
I like tacking on stubs of precut 30awg wire if signal integrity isn't an issue. It's much easier to manage multiple probes and you aren't constantly worried about losing contact. With problematic packages that can't be probed directly you can tack a wire onto a nearby resistor or slip it into a via if you planned ahead and made them big enough. Then you're ready to securely hook on with a scope probe or grabber clip.
That looks like a pretty nifty idea for a 3D printable community design. I'm picturing either copper or aluminum wire for the adjustable-shape part, a spring for the active holding force, and a vertical scope probe for the lead length issues discussed in the article.
I also want a solution, but I think the measurement is far from straightforward. So far what I've found are (Disclaimer: I'm a sysadmin, not a qualified hardware engineer, it's just for fun):
* Using a Vector Network Analyzer and do a S22 (S11 on a two-port VNA) measurement directly at the output driver - it's exactly what we want by definition. But the measurement is not passive and you are connecting the output of the VNA to a "hot" output driver, guess what could possibly go wrong? Both the chip and the VNA can be damaged if it's not done properly. It's quite a headache for testing RF amplifiers.
Even if the setup is a success, everything else connected to the output driver can distort the measurement. But I believe it's easier on low power logic circuits. You add a DC-block to the VNA, get a test fixture and do Open/Short/Load calibration, then solder the output driver on the fixture and turn it on, drive the output high, measure it, drive the output low, measure it again. But it certainly needs to be done on a test fixture in a controlled setting, not an actual live board. I'm not sure if the approach I just described can work yet, I'll do some experiments in the future.
* Weng-Yew Chang, et al. proposed output impedance measurement by current injection using two reproposed wideband oscilloscope current probe (Tektronix CT-6, $1000) and a standard VNA. You hook both probes at the output side of the driver, connect them to the VNA, and do your measurements as usual [0]. Apparently it's already the standard practice for measuring DC power supplies (e.g. see PicoTest Signal Injector manual [1]), and they simply applied the same trick on output driver.
* Richard Allred proposed a method at DesignCon 2014 [2]: First, get a multi-GHz oscilloscope, get a Time Domain Reflectrometer, do a lot of TDR measurements while the board is unpowered. Then, power on the board, measure the real output signal on your oscilloscope. Next, solder a shunt resistor at the output, repeat the measurement. Finally, do a lot of linear algebra on the measured waveform using previous TDR data to remove the influence ("de-embedded") of the interconnection (cable, board) from the signal, and do more algebra to get the complex impedance of the output driver on the Smith Chart.
Expensive equipment and a considerable amount of work. But Allred's work is used on 10 GHz+ systems, and it's designed to be applied directly on the board, without the use of a test fixture, and avoiding a that can potentially upset the output.
https://circuitcellar.com/wp-content/uploads/2020/03/343-Lac...
although these 13 words help it get to 1000:
"An example of a quite exotic old probe, a 40-kV rated Tektronix P6015"