When filing the bug, I uploaded [[:File:Dombron valflaggning 2025.jpg]] and created [[:c:Category:Category:2025 elections in Finland|]]. That's a typical example. Before uploading I could have thought only of the bridge, then realising the flags and election posters making it relevant also for the non-existing election category (in this case there was another existing election category, but at the time the photo was taken, there probably was none).

I assume that backing out of the UploadWizard session makes one lose all info one has added. I don't know of any way to save that information in a way that allows easy re-adding of it in a new session. Just typing in the category name (or copying and pasting it) is quite easy once you are aware of the problem, but for somebody not aware of the issue, the wizard not finding the category (and redlinking it if entered manually) is very confusing.
I understand that the wizard cannot represent the category state live, but there should be a way to have it refresh its cache.
The situation is common: you have a couple of images of something that isn't represented on Commons from before, so you want to create the category or categories. You might not have checked the categories before, or you realise that also some other category would be appropriate. You might realise the need when you already have added detailed descriptions of many files.
One solution would be to leave the missing categories out in the wizard and add them to the files later, but for a dozen files just checking the categories box and clicking "copy to all files" (or whatever it's called) in the wizard is much less time-consuming.

Whether the page exists or not is not that important. If it exists, a link to the page should naturally be given, if not, we want the redlink. With some neutral wording, MediaSearch does not need to know whether the page exists. It could e.g. tell "You searched for [link]."