After more investigation I found the oscilloscope is working as it should.
The value received from the interrupt is around 6788 which is smaller than the 30000 trigger in the code I posted before.
I changed the code to trigger on 3000 and now it logs values.
I tested this without the pinchangeint library and 4 RC channels and I do not see the peeks.
I tested this with the pinchangeint library and 3 RC channels and I see the peeks.
So my conclusion is that the peeks are caused by the overhead of code. This makes me think that some RC peek down signals are delayed which causes values bigger than 3000 are received. I don't think a RC signals are missed because that would give values in the area of multiples of 20000. So I expect there is some delay somewhere caused by pinchangeint library or a combination of my lib and pinchangeint.
As to the solution: I think ignoring the value is safe because I can only get this value when I move to a certain position (a bit above but close to center) when I get out of the peak it continues as if the peek was never there.
It may be related to my RC receiver who may fire all the channels at the same time making that similar values on multiple channels are hard to detect because they all trigger at the same time. I haven't done any physical measurements on the behavior of my receiver to confirm this but given it's intended usage it seems logical all channels would fire at nearly the same time.
Best regards
Jantje