I'd consider myself 2 but I don't think it has to be a matter of what's ideal but more a matter of the apps use case / amount of value that use case is regularly moving (what's valuable to one person is not to another, so hard to nail this). If I lose a few thousands sats on SN idc - would I store anything more than $100? Probably not.

I think custodial platforms could do more to educate and nudge users into the self-custodial direction though as it is ideal if you want to store decent sums in bitcoin and likely has some regulatory benefits if you don't want to be considered a custodian down the line.

An example of how this could be done is limits on how much sats can be stored on SN and educate users about using a self-custody app instead when they get close to or hit that limit (would be great if you could tip on SN using a self-custodial app like Breez too!). You could also have a cap on tip amounts, say higher than $50, to receive a tip the receiver can only receive this with a self-custody app connected to SN. This is a bit of a happy medium between 2 and 3 and could keep SN in the lower bounds of custodianship which could have some regulatory benefits (I'm no lawyer though).

Use un-memorable passphrase and store elsewhere, pseudo-multisig without extra costs / signing friction

100% this, one single hardware wallet with a passphrase is all one needs.

Need P2P apps like Keet asap https://keet.io/

Scales payments, not users.

proof of reserves is a scam, do not trust it

TL;DR You still need on-chain transactions to open channels.

Scaling users, it only scales transactions currently.

Listing grant amounts could be one option. Few people here seem unclear about that and may prevent them from applying.

https://qala.dev is another

Here is a link to the specs development: https://balls.dev/

TL;DR: Headless signers allow users to get the UX benefits of collaborative custody whilst maintaining better privacy, security, and interoperability.