Contradicts:
You have to be as specific as possible, because you don’t have any founded reason to take action at the moment you find some (especially already well-known) behaviour disruptive for your service.
For example, current (outdated) rule of only 1 node per IP easily could be enforced by denying any concurrent connection from the same IP. Can be mentioned in the ToS, but will be self-explanatory at the moment users try to circumvent this rule.
Rules like no more than one IP per physical location (or something like that), one disk/CPU per node, etc. need specifically to be in ToS if you find them important (which I doubt to a certain extent), as long as you don’t have specific and easy to implement ways to enforce them. But you could also list them in a separate list of prerequisites for nodes, which is updated separately and more frequently than the ToS itself; as long as you refer to that list in the ToS and make sure you take the effort to communicate significant changes to your node operators.
The “/24-rule” (which isn’t a rule, see below) is one of the latter, especially because it’s actually a quite raw proxy for location and not easily enforced. As long as we accept the fact VPN’s are being used (also for understandable/acceptable reasons as GC-NAT in my case), you have to be specific on the fact that only one IP per location is permitted. It’s a way the network may become a bit more unreliable, although this shouldn’t be overestimated (see also calculations in the excel sheet). So it’s a precondition for the reliability of the network, not something like “how the network works”, since the network is still working although we assume many people are finding ways to make more IPs bound to one physical location.
Strictly spoken, it’s also not circumventing a rule, because not stated in any legal document. Opposed to, as already argued before, for example the practice of running multiple nodes behind one IP which is explicitly condoned in the manual. So strictly spoken, node operators using a different VPN for every node they start are actually living up to current ToS (and not to be mentioned “cheaters” or whatever subversive language can be used).
So the reason why we’re having this discussion, isn’t the behavior of some node operators (I don’t know whether we even have an estimation how big the problem actually is), but the fact that significant chances in preconditions of the network weren’t reflected in the ToS. So actually legislative/juridical neglect from the Storj-company, which they should account for instead of those some (or many?) node operator.