I agree that inability to achieve consensus in direction has been a major detriment to Bitcoin, plaguing it for years now. There has likely been hundreds of thousands of man hours wasted, developing implementations and code that ended up being scrapped.

When a group of nodes start running a different version of the software, this creates a fork with the exact same transaction history and balances that existed prior to the fork, but moving forward each would operate independently. How miners play their part in deciding which is the predominant blockchain, is by processing transactions; because if there were no miners, transactions would not be processed. Less supported blockchains are also much more susceptible to a 51% attack, especially if there are some large players or groups with massive hashing power capabilities in the given algorithm.

Theoretically, the majority of nodes could be running a version of Bitcoin with larger block sizes, but the majority of miners may still be more inclined to run the original and continue to process full blocks because there are more transaction fees (1000+tx/block @ $30 average tx fee).

When it comes down to it, even the little guy can still play a part in deciding the future direction of Bitcoin. On the Bitcoin network, a node is a node. Given that there are less than 12,000 nodes at this moment (most in data centers), a collaborative effort could still easily be mounted by ordinary people to implement new code more conducive to scaling Bitcoin. But if the miners are not incentivized enough to make the switch, does this really solve any problems? This argument has been going on for what seems like an eternity in the industry, and no ground has been made to break the stalemate.