S3 doesn't really have folders, only single files grouped according to / characters contained in their keys.
However when clicking a folder and selecting Delete from the actions, I remember Cyberduck would recursively delete all contained keys.
Now instead it seems to consider the folder as a file in its own right, failing:
HEAD /23435/ HTTP/1.1
Date: Fri, 06 Oct 2017 16:29:02 GMT
User-Agent: Cyberduck/18.104.22.168405 (Mac OS X/10.10.5) (x86_64)
HTTP/1.1 404 Not Found
Date: Fri, 06 Oct 2017 16:29:01 GMT
The workaround is to go inside the folder and ,manually delete all the files, but that shouldn't be necessary.
Version 6.2.6 on Mac OS X.
I'm dealing with this issue as well. We have versioning on buckets, and a "directory" gets "renamed" or deleted yet it continues to show up in the Cyberduck interface. You try to delete it and it gives you an error. It does not show up in the listing if you use the s3 cli. It doesn't show up in the aws dashboard unless you show versions.
I would expect if you had "show hidden files" turned off this should at least be hidden, and if hidden files was enabled it should be displayed differently to indicate it isn't actually there. It causes a lot of confusion with multiple people accessing a bucket as they are re-uploading things or placing them with previously deleted paths.