Open theremin v3 loses calibration

About the problem of temperature drift, I'm in C2 3.9.10 cog grade capacitors are used. It seems that while they have high precision, they also have lower temperature drift value. They can remain stable after being adjusted 3-5 minutes after startup. By the way, an irrelevant topic. I saw the release of V4. It seems that it just integrated Arduino into the main body on the basis of v3. Is there any improvement in his performance relative to V3? Or what are the hardware optimizations and modifications of V4?

Changements are, as you said, the integration of the AVR CPU on the board which gives additional IO pins and which among others allows to output a GATE signal, and an additional 2-channel DAC, so that pitch and volume CV can be brought out simultaneously. There are now two 1/8" stereo plugs, one for audio (left channel) and GATE (right channel), and the other for pitch CV (left channel) and volume CV (right channel). Instead of the Arduino's rather bulky power and USB plugs, there is a single USB-C plug, now. The RF front-ends (oscillators etc.) are unchanged. The firmware additions consist of generating the pitch and volume CVs and the GATE signal. Some other parts have been code-optimised to create space for the new functions, but without changing the sound or playability. All the other stuff is again unchanged compared to the v3 HD audio version.

Just by curiosity.

On V4, the same quartz is used for picth reference oscillator, sample clock and Atmega328P clock.
Probably no impact on thermal drift, but maybe it has some other positive impact, hasn't it ?
(I remenber that the FAQ of V3 had a remark about poor arduino clones' 16 MHz clock)

The positive impact is just saving PCB Estate and thus leaving space for the AVR CPU and the third DAC. It would have been a waste to put more than 1 Xtal for the same master frequencies.
The problem with poor Arduino clones, using a less precise ceramic resonator instead of an Xtal is thus eliminated, too.