I swapped the 1nf capacitor now for a 680pf and the values read by the ESP32 seems to be fine. The Corsair fan reaches right around 2000RPM and the noctua one about 1600RPM, both values are the manufacturer specifications for the maximum speed. I can't spot any difference on the oscilloscope when us...

Adding the 1nf capacitor did indeed help. The values seem quite realistic but a bit high now. I read 3000rpm ok my 400-2000 rpm fan, while the slower noctua one looked fine.
I will try to use a slightly higher value capacitor tomorrow and get some new measurements with the capacitors.

I am happy to use those additional 8 resistors, I am just not sure if it is the correct solution,

What exactly do you mean with scope readings? Did you just miss my post with the pictures or would you want to see something else?

I discovered that using a longer cable from the pull up resistor to the reading pin seems to bring correct readings. Because of this I tried using a 330 Ohm Resistor instead of the longer cable which also seemed to work... My goal is to turn this into a PCB. Would I need to add those resistors there...

I got a proper oscilloscope now so here are some readings

100PWM 255PWM PWM Cable unplugged The measurings have been taken with channel 1 connected to the Pulled up tacho signal (where the cable would go to the ESP32)

Yes it is

I now used a arduino nano as cheap oscilloscope to look at what I get

Here is the screenshot of the result. (Detected frequency is fluctuating between 24.4 and 25.6 Hz, forgot to turn it on in the screenshot. Fan should be around 800 RPM)
Does this show anything related to my problem?

I tinkered around some more and found out I can in fact get correct speed readings for the fan if I use a seconds jumper cable on the tacho line and touch/thumbe around the two jumper cables. The effect gets lost when I crank up the fan to more than 150 (out of 255). Is there a more practical soluti...

Esp_sprite I too think that will reveal the cause which is that the signal while in essence is correct is noisy or has bounce. Without a scope it could be tested by rotating very slowly to by hand an reporting the rpm. Tom Meyers As I wrote in my previous (not yet published by moderators) post, I d...

Unfortunately I don't have that gear. All I have to measue is a digital multimeter.
But when used in my pc, both fans seems to work correctly.

Maybe ordering some cheap oscilloscope or asking to use one in the office could be possible though...