This problem is overstated IMO. Choosing the one with the most validators is closely analogous to choosing the chain with the most work i.e. the one with the most hash power.

And with algorithmic difficulty adjustment in PoW, you don't pick the chain with the headers with the most 0s preceding, you pick the longest chain. A longer chain with a sudden difficulty adjustment is an obvious invalid chain, but I digress

The bigger problems for me are the nothing st stake problem. Why pick one chain when you're new? Just pick both, it costs virtually nothing, and when you figure out which one is more profitable to you, just let the stake in the other one get slashed. Validators can help an attack on the network just in case it becomes profitable, and if it doesn't they lose nothing, because they incur no ongoing costs in the real world to do so.