There should be an F2L trainer somewhere(please someone else post the link) and after a session it can tell you which F2L cases are slow for you. These are the cases you want to look for efficient solutions and learn algorithms for.
You can do this multiple times throughout your timeline, just let time and solves pass in between sessions(maybe 2 close in a row then 2 weeks pause).
Ideally, one starts off learning intuitive F2L, and as they get faster, they learn more and more algorithms for difficult cases, EO cases or multislotting cases.