I agree that there should be something to prevent replay attacks on the protocol level, both for the exchanges' sake, and for the protection of the end user.
Used properly, the block size itself can be leveraged to disassociate coins between chains.
I can send a transaction (or a chain of them) which would take ages to confirm on the existing chain, yet on a big block chain it should confirm in the next block. I then RBF on the existing chain to a different address under my control. Voila, easy disassociation.
There is also mining reward "taint", although you have to actually get some of it to use it. Once you've tainted your coins though, they can be used to taint others, and so on.
Yes, exchanges (and users) could send some of their BTC to one address on BU, have that transaction confirm, then double spend that transaction spending those blockstream coins to a second address with a high enough fee for it to confirm, and subsequently have each transaction be dependent on each of those non-compatible transactions . Unfortunately that could turn out to be very expensive and impractical, especially if many people are trying to do this at nearly the same time, especially considering that many exchange customers will not like delays in processing withdrawals.
If there was something in the BU protocol that makes a BU transaction not compatible with a blockstream coin transaction, then this would solve the problem.