I just connected the AVR to the electricity and with exactly the same connections than before... and it worked.
Congrats! Maybe the AVR has some type of an EDID buffer, which was cleaned now.
Make some more tests, and then mark this thread as solved. I don't think it's an LE issue.
The AVR probably had bad EDID data (containing a video mode) when connector detection failed. Dev talk:
Quoteexcept for the second HDMI / AVR (which looks like a broken EDID and all modes except 720x576i filtered out) the kernel log looks fine.