I agree that keeping files in an AntConc-source folder makes sense. That indeed is part of my workflow.
Unfortunately, though, rebuilding every time I add a file there is time-consuming when big pdf files are involved, because of conversion time. I'm guessing that AntConcConverter is working under the hood to do that, because corpus rebuilding time is very similar to AntConcConverter processing time.
Since corpus building is amazingly fast with txt files, one solution would be to convert big pdfs to txt manually, store in the AntConc-source folder, and rebuild my corpus with a mixture of txt and pdf files.
It would be a pity to have to do that, though. The ability to add files on the fly to an existing corpus would avoid a lot of this sort of manual work.