For the purposes of the OP, I'd say it's likely you are correct in the assertion that it's mainly game theory keeping that part in check. But that shouldn't in any way be interpreted to mean we don't still need a healthy node-count.
We are arguing on the basic premise that having a healthy full-node count is a desirable situation or not. Your argument is that full nodes are useless in terms of keeping miners in check because the game theory will keep the rational miners in check. Those of us who think that full nodes aren't completely useless say so because of two reasons:
1. Bitcoin network is equally a network of users/ businesses running full nodes that validate and propagate well-formed transactions, which of course is the only thing they can watch out for and not double spends. (You said in reply to @Wind_Fury that block propagation is not an issue. Is it because in a hypothetical 'miner-collusion scenario' propagation will any way happen between their own full-nodes over relay networks like FIBRE ?)
2. Having this network of full nodes is the best way to remain vigilant. If we depend on, say, a few guardian nodes and then media to keep the game theory going then you are again dependent on centralized authorities as DooMad had been saying.
Keeping the ability to maintain and run full nodes should always remain a low-barrier activity not just because of fear of miner collusion/ attacks but because a lot of properties of bitcoin arise from it, namely decentralization and being your own bank, even in the D-day situations, even if media becomes dishonest, even if a concerted attack targets miners in some jurisdictions rendering the miners disconnected. If we just accept that all of this is paranoia and in reality, such attacks are not possible then of course there is no need for full nodes.
couples of hundreds
Home miners plus full nodes is not a backup plan for D-day, it is the original plan which failed in 2011-2012 after
pooling pressure falw showed its power and pools started to dominate the mining scene of bitcoin. Since then we are running plan B: Centralized mining, no home miner, less full nodes AND game theory.
Thanks for the link to your topic on PoCW. It aims to solve the problem of pooling because i guess you do believe that centralized pools can coordinate and they reduce decentralization.(??) Yet, if miners are to be trusted and no full-nodes are needed then doesn't that create the same problem?? Huge mining farms can coordinate. Instead of debating about need of full-nodes, i will go and read about these proposed scaling solutions you think are being ignored due to the love of full-nodes..
I think there is a deep, deep, misunderstanding: Actually, I am the one who is in favor of having millions of full nodes and the guys like @DooMAD and @Wind_FURY are the ones who are against it, practically!

How? Let's dive in:
There is a very efficient, secure way of running a bitcoin full node on commodity mobile devices very efficiently in economic mode, called UTXO commitment. According to this model miners would have to commit to the UTXO (state of the bitcoin machine) by somehow including its hash in their blocks and a light full-node could besides synchronizing headers like a usual SPV wallet continue to downloading the full state of the machine as of latest couples of hundreds of blocks and act like an ordinary pruned full node thereafter. Right?
For this to happen, you need a simple soft fork which is not getting support, why? Because people are paranoid about miners being susceptible to commit to bad UTXOs for say 500 blocks!
Are we clear? This full-node, full-node propaganda is totally fake and misleading. It is not a campaign in favor of empowering bitcoin users by giving them an opportunity to run a full node, it is just a trick to relying less and less on miners in a baseless and paranoid way: Who in the hell can imagine a scenario in which miners are colliding for 3-4 days to inject a few false UTXOS in bitcoin blockchain? A paranoid or a politician.