Why can't ETH do consensus voting every 3 blocks instead of every block?

From what I understand, the reason there is a 12s interval between each block is because of consensus voting, and the reason each block has a gas limit is because of sudden bursts of network and CPU/RAM usage.

What if we still do consensus voting every 12 seconds, but every consensus voting includes 3 blocks? So the block time is 4s but consensus voting only happens every 12s. A 4s separation between sudden bursts of high load seems more than enough.

I heard delayed execution is on the way. It basically spreads block processing across the entire 12s, so there will be no more sudden bursts. It should allow us to have very large blocks, but it is still years away. It seems much easier to just make consensus voting happen every 3 or 10 blocks instead of every block.

I'm pretty sure there is a good reason not to do this, otherwise it would have been implemented already, but I don't know what it is.