The problem is that many of these setups are put behind essentially consumer connections, but with a better support contract or simply fitted with a 5G modem and a SIM card from the cell provider with the cheapest plan.
Building a WAN isn't magic, but it does require working with (competent) ISPs and paying the higher cost.
If it's a lot, maybe the manufacturers should be compelled to provide security patches if remote control is required to operate their products. For the sake of national security, of course.
Yes you might need to be able to centrally monitor and manage things but that should be happening on a private network.
How exactly would a more obscure (or even “analog”) encoding help? This seems orthogonal to reducing surface complexity and remotely executable command subsets.
The idea is not to solve everything with one simple trick, but to reduce the attack surface as much as possible in hardware. If you are dumb enough to make your 4 messages something like off, on, hard reset, random dangerous maintenance protocol, then it does nothing. If instead it encodes 4 useful and typically used states that can't destroy your system, then you have massively reduced the requirement for onsite, and accepted that more serious issues require onsite control.
It's not an easy balance to strike, though; the less capable the remote interface is, the less useful it is. "Only completely harmless commands" might simply not be good enough. For example, is "deactivate electricity output of this solar panel" (e.g. because market conditions make it unprofitable to supply at a given instance) harmless or not? What about "increase supplied frequency by 0.01 Hz to stabilize the grid"?