Earlier you suggested 2-3 or 3-5 multisig. Even though I understand that it's the common use, I don't really understand why not using 3-4 for example. Is there any other downside apart from the obvious, that losing 2 seed phrases would lead you to losing all of your funds?
There's no reason you can't use 3-of-4 if that's what best fits your threat model. I simply used 2-of-3 and 3-of-5 as examples, as they are the most common. There's no downside to 3-of-4 when compared to 3-of-5, other than less redundancy and resistance to loss as you point out.
Finally, is there anything wrong if I used a private key both for a singlesig wallet and as a cosigner in a multisig wallet? I suppose this question must have already been asked somewhere in the forum
This is certainly possible, but there are a few issues. The first is the obvious security and privacy implications of reusing the same private key across multiple wallets. The less obvious issue is the derivation path issue. If you pull an individual private key from your multi-sig wallet and use it to generate a single-sig address, then you can run in to problems restoring that address from your seed phrase since that address will be at a multi-sig derivation path, rather than one of the common single-sig paths.
A better option would be to reuse the same seed phrase to generate both a single-sig wallet at m/84'/0'/0', as well as one cosigner of a multi-sig wallet at m/48'/0'/0'/2'. Although the seed phrase is the same, the private keys will be completely different.