Everything you do in Bitcoin is voluntary

nostr:nevent1qqsv5gw49lw60vcgnk6uhszf6cw5ndd49xermkzkrlr7xutlmgff9tspzpmhxue69uhkummnw3ezuamfdejsygqpphcvjj87n2656t9hafpql7sg64u43xqmd6ngaqa25l4jm5lstgpsgqqqqqqsepkttc

in case of soft fork, new TXs will be valid for old nodes too

So what version of bitcoin core do you think we should be running right now?

If there were a malicious soft fork that still follows current consensus, has majority of hash rate, and many don't agree with it, some nodes would have to be a bit more involved.

Let's say Foundry, Antpool & F2pool all have dominion over the machines running on their pools (which they don't, but let's say they do). They collude (well over 50% of hash), and say they're not including txns that spend from addresses starting with bc1. That'd be allowed by all our nodes, since it's not breaking any existing consensus rules, yet the economic nodes would be highly incentivized to scrap that new more restrictive rule.

There would already be fat fee txns in the mempool that spend from bc1 addresses, and minority miners wanting to include them in blocks. Economic nodes could enter the command "invalidateblock" and use the hash of the first block of that lousy softfork. Once a block is mined by the minority miners that includes a bc1 txn, it would be included in the first block of the chain the nodes want and ignored by the malicious miners. The malicious majority miners could keep mining on their softfork chain, but it includes that block that you invalidated, so you don't care about it. To keep with consensus (and make money instead of bleeding it) they'd have to abandon their soft fork and jump on your chain. You don't ever have to worry about their chain overtaking yours, since it includes that invalid block, however deep down it is.

You could slip into a coma, come out of it 10 years later, plug your node in, and it'll give you the consensus history like it always does.