Recently, a bunch of narratives about parallel/sharding have gained popularity again, with testnet incentives, point expectations flooding the screens, and everyone in the group guessing whether the mainnet will issue tokens.

While it's lively, what I care more about is: are you risking your assets by putting them into some contract, signing certain authorizations, and having the illusion of “I can always revoke it”?

To put it simply, no matter how much performance is hyped up, security still boils down to the three old principles: who holds the permissions, whether it can be upgraded, whether there are blacklists/freeze options, and how you exit if something goes wrong.
Especially with cross-shard/cross-chain bridges and routing, the biggest concern is “temporary admin + upgradable proxy.” Once the logic is changed, you won’t even have time to react.
My current approach is to avoid using real assets if I don’t have to, and if I do interact, I first reduce the authorization limits and go through the exit process myself…
Anyway, don’t pin your hopes on “revoking it later.”