I made some more work on firmware, wrote routines to support both voltage and resistance measurements and noted some facts:

1, Reference resistors R6 and R7 in schematics https://github.com/jaromir-sukuba/micro_progmeter/blob/master/hw/6c.sch.pdf should be decreased, something like one order of magnitude. When MUX3 signal is active and R7 engaged, measurement range is up to roughly 500kOhm, but the device is lacking precision under 500Ohm. On second range (MUX2/R6), the measurement range is well above 10Mohm - this is higher than expected. Something like 47k for R6 and 1k for R7 should move the measurement ranges lower into more useful range.

2, Even with no calibration/correction, the gain error and offset is quite acceptable, though I'm going to emply some kind of calibration.

3, The floating point routines are FLASH hungry, especially on SDCC with "not that finely optimized" libraries.

Above that all, it starts to look like a test gear, kind of.

Here are displayed debug values with measured resistance in last line. Close enough, for uncalibrated device.

Bunch of resistors tested with this one. Those cheap carbon resistors are just for basic sanity testing, old precise resistor decade is still waiting for its job.