*facepalm* I accidentally deleted ETFbitcoin's post above my reply, it's content is preserved in my quote (I'm not sure that it's all of it, unfortunately)... the quote/edit/delete buttons are right next to each other, and I don't use BCT much of late. I'll send him a PM to repost it, and I'll fix the thread order.
1 and 2 are correct, 3 is incorrect because it's talking about a different thing.
The post you've linked to is saying, effectively, "The rules for witness type v0 further restrict the scriptpubkey to have exactly 20 or 32 bytes of payload, since that's the only sizes v0 can use. With v1, instead, if the size isn't exactly 32 bytes (the only size that v1 can use) then the v1 rules aren't applied and those are treated the same as the other versions or v1 before the change: nothing is done, only the same validity rules apply as apply for all scriptpubkeys" -- none of this has anything to do with data in the witnesses.
So for a scriptpubkey to get v1 treatment it must have v1 and exactly the right sized payload. This leaves room e.g. to create a v1a that (for example) uses 33 bytes to add the sign back to the scriptpubkey or introduces 448 bit ECC without having to burn up an additional witness version number.
This consensus rule design doesn't improve people's ability to spam using large scriptpubkeys because the same large sizes are valid for non-segwit transactions and those couldn't be made consensus invalid without potentially destroying the bitcoins of someone who happened to have encumbered them with an nlocktime release that involves some large script.
Maybe the invalidity risk there is fringe enough that it could still be argued for, I think not, since the underlying attack that you'd be trying to stop doesn't really care *how* the data is encoded, closing off one will just cause them to use others-- but regardless, going around hobbling bitcoin's consensus rules to try to play wack a mole against some spam vector is a pretty *distinct* concern, and should be handled if at all separately and holistically. Could you imagine if BIP 342 had an "oh yeah, and also all scriptpubkey's-- not just v1 but all unused witness types and non-segwit too are now limited to X bytes, hope you don't have any funds using long scripts that you can't move before BIP 342 gets activated!"? lol