I have no horse in this race, don't dive a CCR.
However, I think it's slightly disingenuous to say the cells "work pretty well" when the context that makes them do so is using at least three at any given moment and having an elaborate set of voting logic and processes to detect and ignore the non-working cells as part of everyday CCR diving. That the end result works isn't a testament to reliable cells, it's a result of the layers of redundancy and workarounds built on top to handle the fact that they are unreliable.
There was a well-publicised cell-related death a bunch of years back (two cells went bad on the same dive), but I think the consensus was that it was human factors (using way too old cells, ignoring alarms, etc) that really caused the accident.