I think the issue here is that it isn't really "push" they want to do. From what I can tell, they want to do is simply "run all the time so we can keep a connection open to receive new messages immediately". Which, to be fair, is pretty similar to what Apple does for their push notifications.

However, allowing this would be a bad move. It'd mean that rather than once centralized connection that knows about the device's network state and can (at least try to) preserver battery life, you'd then have as many as one per each app, always open. (Once Sparrow does this, then why could any other app not?)

The real issue here is that, from their responses, it seems that Google doesn't have some sort of authentication method like that could be used to get access to just enough information to pass the push data to Apple — or, at least, not one (like OAuth) that doesn't need them to remotely store a username and password. If there was a way for them to get some sort of useless "token" to get the notifications themselves and pass them to Apple, I don't think the security issues would be anywhere near as severe (especially if it was an opt-in service). But I don't know if Google does that, and their responses don't seem encouraging that something like this would be possible.