> This is "broken" in the sense that it makes assumptions about how the
> "multiple values" are implemented.

Yes, though we've had that implementation for decades and we're not likely going to add true multiple values to elisp.

> We do know that `ibuffer-split-list` returns a list, so there's no point
> going through CL's multiple-values simulation.  We should instead treat
> it as a list and extract the elements with `pcase-let`
> (or `cl-destructuring-bind`).

I tried to do a minimal fix so close to branching for 27, but I can do that if you want.