I have made some progress, I will describe it here for documentation purposes, maybe it can help others that are battling interference.
I decided to try another approach and construct an antenna that is less prone to local interference. It is well known that a horizontal dipole over a conducting ground screen has a deep null towards the horizon, rendering it less sensitive to local interference. Furthermore the ground screen blocks any interference that comes from below the antenna, so if the antenna is mounted higher than the noise source it should not be picked up.
To test this hypothesis I put about 2x3m of chicken wire on a flat part of the roof and mounted a quick and dirty single dipole about 3/8 wl over it and scheduled some observations. The results are not bad, this observation is from this morning when the PV-system was just switching on:
Observation 248381: Fox-1A using dipole-and-screen antenna, PV-system switching on and off
You can still clearly see the noise floor rising when the PV-system switches on, but the effect is not as pronounced as in earlier observations. It could well be that the interference is now picked up by the cheap RG-58 feed line that I’m using in this quick 'n dirty setup. I will try eliminating the feed line by mounting the SDR at the dipole, to see if that helps. This will definitely help getting the SNR up. Of course this antenna is not very suitable for low passes because of the null at the horizon, but anything over 20 degrees should be OK.
To be continued.