Which database are you using? Which specific show are we talking about? Can you provide a sample file path?

If you fetch the episode information for a given series every ~6 days, then you will always get the latest episode information from the database at that point in time, so you should not have an issue with weekly shows. Especially if you process files ~1 day after release.

If you fetch fresh episode list information today, and then try again a day later, you will instantly get stale episode list information directly from the cache on the second day. However, if you try again a 5+ days later then you will again get fresh episode list information.

If we are talking about a series that airs weekly, then the easiest solution is processing the new file once per week
(this guarantees that we're always working with fresh episode information) one day after release
(this probably ensures that somebody has replaced TBA with the episode title already).
EDIT:
-clear-cache isn't really a solution to the TBA problem because you will still get files incorrectly titled TBA if the database says TBA at the time.
-clear-cache is useful if you then want to fix those TBA episode titles a day later, and don't want to wait for half a week.
-clear-cache is
not necessary if you just wait a few days before re-processing the file.

A good solution would process the files correctly on first try, and so we'll want to wait for others to enter the missing episode titles, and process the files after that. If you process files manually, then it's just a matter of doing it a few hours later than usual
(instead of immediately, getting TBA, and then being locked into TBA for the next 3-4 days due to caching). If you have an automated setup, then things may or may not be easy depending on what you're doing. If you are using the
amc script then you can use
--def minFileAge=1 to process only files that are older than 1 day already. This approach works well if you process files on a schedule every day. If
filebot is called on newly completed files directly, then you'd have to find a way of deferring that call until later, i.e.
sleep 86400; filebot ... would be a most basic solution for that.