On equipment, the honest answer is that there is no plug-and-play sub-THz sounder — everything above 100 GHz is really a VNA plus frequency extenders, or a wideband arbitrary waveform setup with upconverters, and the measurement is only as good as the phase stability of that chain. I have used a VNA with D-band and G-band extender heads, and the two things that mattered most were keeping the extender heads and the waveguide runs thermally stable and minimizing the number of waveguide joints between the head and the antenna. Every extra joint is a repeatability risk.
To NVMe_Dev_Adam's temperature question: yes, it is very sensitive, and that is exactly the trap LunaWei flagged. The multiplier chain in the extender head drifts with temperature, and the phase of a meter of waveguide or even a phase-stable cable moves by degrees per degree Celsius at these frequencies. In practice I have seen a few degrees of phase drift over a 5 C room-temperature swing, which is enough to invent a false multipath component or to shift a real one. The amplitude is more forgiving; the phase is not.
What works for me: warm the whole chain up for at least an hour before calibrating, and then calibrate immediately before the measurement run, with the heads already mounted in their final positions. I keep a short back-to-back thru measurement as a repeatability check and re-run it after every sweep — if the thru phase has moved more than a degree or two, I do not trust the channel data from that sweep. I also log the room temperature and the extender head temperature alongside every capture, so a drift shows up in the metadata rather than as a mysterious new path.
For the channel itself, the other check is dynamic range. If the noise floor is within 10 dB of your weakest real path, you will start calling noise multipath. I always record a blocked-antenna noise capture first, so the threshold is measured, not guessed. If you share your band and your VNA-plus-extender combination, I can compare notes on what repeatability you should expect.