I don't think the fact that not all nodes are mining nodes changes the fundamental premise. In fact it strengthens it as there are more validation nodes.
No, you missed my point. I thought it was more evident.
Non-mining nodes are not incentivized the way Miners Nodes are.
The point being:
both can have opposite votes now (two sides), instead of the vote
always tending toward the same incentivized choice by Hardfork.
Now there are more possibilities such as Softforks. (User activated
fork is now another one, but is a new theory which is riskier than
either soft or hard.)
All I'm saying is when you cite Satoshi about Nodes, it was before
the Node split and is entirely in a different context that no longer
exists.
But non mining nodes are the majority of the network. Miners have to produce blocks that follow the network consensus or they get orphaned. Hard fork consensus means nodes update first, miners follow.
That doesn't matter. Miners can fork away without node validators
and the minority hash chain's difficultly will be too high (and may not
adjust in time to prevent possible failure). Meanwhile the Miner nodes
are on a new chain, mining & semi validating away. The minority hash
chain, could in short time, have no hash at all, in theory.
Your statement only applies if there is no malicious forking.
When the two node systems split, this possibility came into existence.
Prior to this, they always moved as one, since there was only one.