You misunderstand my use of "distinguish" there. Even with IPv4, you can already be distinguished from other users when it comes to your identity/fingerprint, because such fingerprinting doesn't come from IP metadata, but from application-layer data. However, filtering based on this requires deeper packet inspection, which is more resource-intensive.
With IPv6, endpoints can distinguish your IP packets from those of other users/households based solely on the IP address. That doesn't increase or decrease your privacy in any way; they still don't know *who*, only *what*. The only thing an endpoint gains from this is the ability to more selectively block/filter packets based on IP address alone. That's good for both you and the endpoint, not bad.
In other words, you have no good reason to think that distinguishing household A from household B, whilst still not knowing anything more specific about A and B, such as their street addresses, is harmful to privacy. Or at least, so far, you have failed to convince me that this actually harms your privacy. Perhaps you could give a concrete example of where this would be the case?