This feels like a misunderstanding of how the syncing process works. You don't only download the last n blocks and assume the current state is accurate. You download the entire chain, validate it, and optionally discard the redundant information. As far as your client is concerned it still verified the entire chain end-to-end. The only difference with pruning is that you can't revalidate the chain end-to-end without redownloading the whole chain, but I'm not sure why you'd want to do that unless your machine was compromised.