A little bit of further checks into the meter... In case someone finds these bits of information useful/interesting...
The chip gets multiple supply "rails", of which most are simply branched from the same 3.52V supply (probably "3.5V"), and one "5V" (4.96V). The 5V has a bigger bypass cap on it, and seems to be stable with no rippling at all times. All the 3.5V supplies have basically same behavior, due to missing the larger local bypass caps and not having much bulk caps anywhere else, either. The 3.5V supply/supplies has/have from "zero" ripple to about 80mVpp ripple (at measurement frequency or multiple of it) depending on the ongoing measurement (what part and at which frequency). Whether that affects accuracy and/or operation would depend on chip's internal PSRR and measurement method/timings. (Considering that other meter variants with the same chip keep those caps, I think it might be better to have them than not.)
I added a 100µF electrolytic at somewhat bad location (where I was able to keep it in contact with one hand while probing with the other), and that reduced the worst case 80mVpp ripples to about 10mVpp. I assume putting multiple 3216 tantalum polymers where they should be, in the unpopulated spots, would drop it even more.
All supplies, and also the adjustable (reference?) voltages have quite a bit of noise on them. However, the measurement setup wasn't optimal, so the noise could be also pickup from the room or the nearby oscilloscope. Anyway, in open case surgery, the noise level was also about 80mVpp (and didn't vary much, although was a bit less at one particular case... could have been also a measurement error, or effect of having hand in different place, or whatnot). Since the case and electronics aren't shielded (other than the ground plane being on the top side of the PCB), I'd assume the noise levels would be similar even with closed case. I was thinking of trying an ad-hoc shielding, but couldn't figure out how to have both easy shielding and easy probing under that shield, so I skipped that.
The measurement voltage seems to be a sinewave biased by +1.80V, amplitude of the sinewave depending on the component and meter's state. The largest waveform I could probe was 1.92Vpp (with a 4.7k resistor, didn't bother to try bigger ones), i.e. from 0.84V to 2.76V. Semiconductors would definitely trip. I have also seen much lower measurement voltage, but that typically happened during "OL" states. (EDIT: I did not test all possible combinations/cases, but the ones I did, both sides of the component are at 1.80V bias, i.e. there is no (significant) DC current through the component.)
On the note of meter "state", I noticed that in the auto-LCR mode the shown result may depend on the way the measurement was started. For example, when checking a capacitor, if scope probes were connected at the time of switching measurement frequency, the meter could end up showing R(s or p) with OL or low value (can't remember which). But if the probes were first detached, then switching frequency, then letting results stabilize, and then finally connect probes back, the meter would happily show and keep showing the correct capacitance value. Not a big thing, IMHO, just to be aware that depending on which order one connects the component and changes options may give unexpected results.