Its not a limit imposed by the watched folder end point.
Usually a stage folder is created when the respective LC service/orchestration picks a batch from Input folder for processing.
Now depending in the LC service your are using (it's pooled instances configuration) and hardware specs, your service might already be running a set no. of instances.
it's probably not able to generate a new service instance to consume the next batch of files, hence the service waits to finish chewing a previous batch.
AS soon as an in-execution batch finishes, that folder goes away from stage, and your service is able to pick the next batch.
Probably that's why you see these symptoms on the watched folder side.