It's also a problem with networks that bill by usage but spread both fixed and usage-variable costs across the expected usage, even if they don't have issues cutting off unused parts of the network; a sharp drop in usage underfunds the costs that aren't variable.

Because that shouldn't be charged after the fact, ideally it should be priced into the water prices as consumers purchase it.

That's assuming pricing scales linearly.

More likely, it takes a flat-ish $x to maintain the infrastructure, and the reduced usage means they can't reach that minimum $x threshold.

To not charge it after the fact, you’d have to recoup costs much faster in the earlier years of the system existence, causing crushing water bills to exist.

It’s not a policy issue, it’s an education issue.