Fair question. Before proceeding to it let's have a comparison between the role of game theory and full nodes (supposedly millions of them) in securing bitcoin:

Full nodes have no power to mitigate the real attacks a hypothetical 51% collided pools can carry out, no matter how many of them are present:

- Short-range chain rewrite attacks are not detectable by full nodes, they just bend the knee to the more difficult chain, forget about resisting against such attacks.

- Long-range attacks are the same as it is just a relative concept.

- Full nodes can't resist censorship/blacklisting attacks against peoples/institutions/nations.

- Full nodes can resist against protocol breach and most importantly double-spending/illegal-inflation attempts very well.

Game-theory addresses all those problems vigorously with the exception of censorship attacks: There is a hole.
It is possible for an aggressive coalition of governments to push for blacklisting groups of people by pools/big-miners and convince other people not to care as long as they feel somewhat/somehow safe (wrongfully). Pools/miners might conclude the negative impact as being tolerable compared to the legal penalties they have to bear otherwise.

Back to your question:
Answer: To use their power for implementing excellent features and solving crucial problems where it is needed.

Right now, bitcoin needs mobile secure wallets, doesn't it? A simple UTXO commitment enhancement could provide the opportunity for having elegant full nodes to be set up in a few minutes on an ordinary low-end mobile. The argument, f it deserves to be called an argument at all, is the vulnerability of this proposal to stupid miners who are susceptible to run a 500 blocks rewrite attack on bitcoin!



Speaking of game theory it guarantees