Jcolombo:
I measured capacitance with the tester at the board, from one side to GND pin header, and with bridge to GND very close to each caps, that was 80 to 82p. It should be 44p (!?)... I think that is confirmed your theory!
Ah, now that is tricky, because if you measure it with the power off, then the characteristics of the MCU chip are quite different and it effectively has back-to-back ("varactor") diodes connected across each pin.
If you measure with the power on, then it is attempting - with varying degrees of success - to oscillate which is very likely to interfere with the measuring function of your meter. To actually measure the capacitances, you would probably need to reprogram the chip to disable this crystal oscillator and remove the crystal itself.