I guess economists would wiggle out of this one by pointing out that, unless price rises happen in non-substitutable staples, people can always switch to other foodstuffs.

They could, but they'd also admit that the switch was being forced and not a matter of choice. Whereas if prices dropped while wages rose, buyers would have increased choice.

I think if you use choice as the bottom-line criterion a lot of the conceptual confusion disappears.

The problem becomes quantifying that choice. And it's possible a simple one-line model may not be the  best way to do that - and also that choice will vary for different groups, and optimising choice for one demographic may lower the choice of another.

As usual it's a political problem, not an econometric one. You can't create good metrics until you decide what you're trying to achieve socially and politically.