Got it, no epochs were left when I set the reject_criteria to the stringent threshold I used, so no epochs were left in the ICA, resulting in that error.
So I think everything can be traced back to the reject_criteria.
For the critera: Why cant I set the flat criteria to -100 microVolt? I actually want to have those criteria: amplitude threshold (-100 to +100 µV) and a joint probability threshold (SD = 5)