This situation is analogous to banking customers (the bitcoin userbase) depending solely on central banks (miners) to uphold the integrity of the system. It doesn't work in the real world and it won't work in bitcoin. Non-mining users and miners have competing incentives. Miners are only concerned with profit -- historical attacks (withholding, double-spend, tx censoring) support that. The only checks on miners' incentives to attack other miners or users are 1) other miners (who might control enough computing power to prevent computing power based attacks) and 2) non-mining nodes (who might control enough nodes to prevent Sybil attacks). Past that, miners are presumed not to be honest (they have clear incentives to be dishonest), and will steal everything they can. Non-mining nodes are therefore essential to keeping miners honest, by making it too expensive or difficult to mount certain attacks against the userbase...
Sure, miners' interests != hodlers interests (though this rabbit hole goes too deep for comfort -- all nodes (wallets) mined in satoshi's model. Contrary to Pierre's claims, the white paper makes no mention of non-mining, non-wallet nodes.
My point is not that miners' interests inevitably coincide with those of the hodlers, but that non-mining nodes (which are not being used as a wallet) add nothing to Bitcoin security. Your wallet (as in the only wallet that's truly trustless, one running a full node) checks for *you* that the miners are following your ruleset.