The definition of hard fork is that it’s a relaxation of the consensus rules; in other words, some blocks are now valid that previously wouldn’t be valid. Because a block with some transactions replaced with a ZKP wouldn’t previously be valid, this is by definition a hard fork.
Whether this would actually end up splitting the network depends on what the trusted committee would do:
The Good
The allegedly leaked messages suggest that such data deletion would only apply to “buried” blocks, i.e. not the newest block. In this case the original transaction data would already be widely disseminated; as long as there are some nodes running the pre-fork version, the data should remain available for newly spun up pre-fork nodes to sync.
The Bad
However, there appears to be nothing stopping the trusted committee from colluding and deleting data this way from a transaction that was never made public, and then getting a miner to include it in a block. In this case the network would split, as pre-fork nodes would never receive the original transaction data needed to verify and accept the block.
The Ugly
The current state-of-the-art ZKP systems for verifying the Bitcoin blockchain (e.g. Raito) are (as far as I know) currently limited to verifying the longest header chain and transaction inclusion; they can’t yet verify script execution. If this proposed fork is to be deployed anytime soon, it seems likely that the ZKP wouldn’t actually fully validate the spending conditions of the transaction inputs, opening up the possibility of the committee colluding and stealing funds.
