Lately I've been looking into IBC and various messaging/bridging solutions, and the more I look, the more I feel that cross-chain is basically "who do you trust."

A transfer from A to B involves not only trusting the two chains themselves but also trusting whether the light client/validation logic is correctly implemented, trusting that relays don't arbitrarily modify packets (although in theory, modifications shouldn't pass), and trusting that the other chain handles timeouts, replays, and other edge cases strictly enough.

So does that mean IBC is definitely safer?
Not necessarily, it just makes the trust assumptions transparent to you.

In fact, many new L1/L2s are launching incentives while pulling TVL, making a spectacle of it.
Long-time users complaining about "mining, liquidity, and selling" is understandable: liquidity flows back and forth, and the risks of bridges and messaging layers are treated as negligible.
My usual approach is to break down the components first: who signs, who proves, who holds custody.
Draw a small schematic diagram; otherwise, I really can't sleep.