Hi Markus,
That's a very good question and I will read everyone's input with a great interest.
One of the faintest features that I imaged with my 105mm telescope and that I am aware of was registering on average 1 photon every four 15-minute subs; signal in the area of interest was approximately 1 ADU above the floor in the final stack, where gain was 0.26e per ADU. Camera's read noise was 4e = about 16 ADU. I would need to measure how much sky glow was in each sub though.
It took about 10 hours of exposure to be able to (fairly) clearly notice this faint signal in a stack. Reaching 50 hours of exposure made this faint signal more distinct, but I was noticing little improvements with added exposure and estimated that a visually significant improvement would require at least another 25 hours of exposure (on top of final 50 hours).
I would venture to suggest that the faintest noticeable (without straining your eyes) signal that can be captured with 10 hours of integration with 3nm filters at f4.5, good QE, 105mm aperture giving 1.60 arcseconds per pixel and with darkish skies (backyard in a small town) is about 1/15th of the read noise.
To illustrate, the above-mentioned signal was in the thin long bows at the top of this image:
https://www.astrobin.com/full/364378/G/