This is a bit inexact.

If there was a mechanism where your node would say "hey, I'm a node, I have X Bitcoins, and I will only accept the Bitcoins of version Y or with feature Z!" (and telling always the truth!) then your hypothesis would be correct. Of course that doesn't exist, and it would also be against the censorship resistance / privacy ethos.

The real power lies in those people that hold and accept Bitcoin, not in those that run a node. Because in the case of the fork, they can choose which chain's coins to accept and which chain's coins would be sold (see ETH/ETC for an example that got really serious with a fork, where Vitalik was able to abuse his extreme holdings of ETH to roll back some transactions). Even those running a SPV wallet and to some extent also those holding / accepting Bitcoins on exchanges or custodial wallets, are more "powerful" in this game than those who only run a node and have no Bitcoins on it.

I think "node signalling", however, is not entirely useless. The evolution of node versions and software packages can give some hints on what the people who hodl and accept Bitcoin are also thinking about certain features, above all the "power users" (who tend to have relatively large holdings). Technically @gmaxwell is correct, this only makes sense really for consensus changes like Segwit. But even in the case of standardness/default values it is one indicator, alongside things like the "sentiment" in social media, which can in the future in theory lead to ideas which would break consensus (see this rumour / idea of the multisig federation hardfork deleting "illegal data"), and then the "sentiment" would become also technically relevant because an UASF would be again possible.

All these indicators can be manipulated and sometimes be misleading, but together they can give you a picture of the reality sometimes.