To define what kind to use. I didn't consider that polluting the kind list was a good idea specifically for an event that might have very different shapes as every state machine state snapshot might have different shapes, and picked the only agnostic kind in nostr for the only replaceable event defined in the spec.
Discussion
Eh, it's not polluting, that's what kinds are for, is to partition the protocol. But I see what you mean, if the schema varies there really is not a lot of interoperability.