the rational is simple, whilst the patches are applied and the instance is in upgrade mode, you don't need other factors muddying the waters.

one of the primary reasons for separating databases and logins to a separate instance is where you have different security requirements. So, yes it is completely valid in your case

better to control this yourself, suspend data movement at the primary, apply patches and resume data movement

dont blindly compress tables and indexes, especially tables that are write intensive. It's more suitable for static tables

generally it's bad practice, this should be carried out in a separate environment