Yes, your understanding is largely correct. This originates from the situation that for each detector on a nanopore array there are 4 pores. I'll explain mux scans and groups, but this is outdated information as now another system is used.
So a MinION FC has 2048 pores with 512 sensors. Not all of these pores will be equally suitable for sequencing. At the beginning of the run originally a mux scan was performed, to test for each sensor which the best pore would be (if any) out of the 4 available. The best pores will be used for sequencing first, the rest will be turned off. That is group 1. After 8 hours the expectation is that many of these pores have degraded/accumulated damage or are not functional for other reasons. Then it is time to swith to group 2. So you change to another set of pores, which in these cases often lead to a temporary increase in yield. The default script used 8 hours, although some used adaptive software to select new pores when the first group started going bad. Also note that during the run no new mux scan was performed: the software did not check if the selection it made in the beginning still made sense. Pores in group 1 got turned off after 8 hours and switched to group 2, without taking their current 'health' into account. Cumulative yield plots get a characteristic bumpy pattern from this. Each bump is less than the previous one, eventually flatting off.
Now that's history, as the software now uses dynamic muxing and will continuously check for the best pore per channel - and as such maximize yield by having as many healthy pores as possible available for sequencing. Cumulative yield plots are now smoother.