One can easily imagine a scenario where miners run overnight when power is cheap, turn their machines off during the day when power is in high demand and expensive (and you'd either lose money by having them on, or you would make less than you would by conserving your hardware and optimizing its usage), and earn a profit overall (while leaving the power producer better off too by letting them sell power that would otherwise be wasted).
Second, there are many better uses for cheap power - one could use it to store energy in a hypothetical future where we use lots of intermittent renewables. You could use it to produce highly energy intensive physical goods - aluminum, hydrogen gas, fertilizer. You could use it for intensive climate engineering with carbon capture.
Almost anything else you can think of would be better than running a proof-of-waste cryptocurrency network.
[1] https://paulbutler.org/2022/the-problem-with-bitcoin-miners/
And is this way - a price floor - and not the way you describe it, because of the economic incentives of miners. They have already paid for these captial intensive mining rigs, and to best turn a profit they must be running at all times. The marginal cost of mining is important, but given the capital costs (incl depreciation of hardware!) you cannot ignore it.
Basically, your explaination is a failure of first order thinking. To a first order approximation, only the marignal cost of mining matters and thus the scenario you describe is true. However, you must include the second and nth order effects of capex to truly match reality.