Most operating systems these days allow setting search. On macOS, I would do cmd+space and search for Wi-Fi. The first result points to all the settings related to Wi-Fi.

Another tip for particularly connecting to Wi-Fi is to head to neverssl.com as first thing after the connection succeeds. Some Wi-Fi providers, such as on airports, are notorious in not showing up the-connect-to-wifi window automatically, but visiting neverssl.com forces that window to show up.

From the OS level view, this sort of problem can be handled by approaches which track state and ‘package’ it. For example, NixOS (or Guix) maintains each OS snapshot separately and switches between them, so it would let you ‘roll back’ your OS to the previous week’s installation by simply switching all the addresses to point at the old snapshot. Since both of them are there, one can then diff them to try out changes one by one, in theory, to isolate the error.

A lot of config errors like this are due to user data in $HOME so it might not help to change the OS, but you can version-control your data (you are backing up your data, right?) and do the same thing: your data is saved, so restore the last week’s version until you figure things out, then you can roll back forwards. (Again, at least in theory. Like most people you’re probably doing remote backups so wouldn’t have the data with you, and of course, your smartphone is so gimped, and so designed for the lowest common denominator, that all of this is a pipe dream. It’ll never give you this sort of control.)

This sort of stuff is why a lot of software development and cloud stuff has moved to Docker and Kubernetes and other stuff slapping on even more layers of abstraction: do everything as code recipes for building an OS+apps+data, and then just ship out entire OS images to VMs and blow them away the moment you don’t need them. Some config change at runtime proved bad? Don’t try to recover from it or figure it out, just kill the bad VM and stamp out a fresh one. (This often works for the same reason ‘crash-only software’ and reboot works: most errors are transient and will not repeat themselves, so ‘have you tried turning it off and then back on’ actually does work.)

along the lines of @gwern’s suggestions, these days lots of OS’s will package the state and allow rollbacks to previous (and possible good) states. Similarly there are ‘safe-mode’ states that disable a lot of extras but leave the nitty gritty (often including networking) to work with. E.g. linux has recovery mode and boot to cli, windows has rollback and safe mode, mac has time machine for rollbacks and a safe mode iiuc. Many of these avoid the bootstrap problem of ‘I could solve this problem if I had network access, but the problem is such that I dont have network access’. A possibly better approach is almost universally taken for extra-monitor setup – there’s a window that pops up saying ‘can you see this window, if not the change you just made that messed up your monitor settings will be undone in X seconds’ although arguably that’s more or less equivalent to rollback.