This isn’t true, you could still mine (poorly) with reduced RAM by recomputing as needed. And I’m not sure what you mean by “clocking-in” a share in a useful timespan. Compromised devices aren’t people, so they don’t care if a particular machine can’t get regular payouts. If the whole network can reliably find a low diff share solutions anywhere it can make money.

It would be horribly inefficient however a network of tens or hundreds of thousands of low resource units can absolutely make money since the costs are zero.