I cannot see how even 1000 sibling directories should create a bottleneck. To me, another typical design mistake seems to have been made by the author namely that all data was transmitted at once, i.e. all directories and subdirectories with files. Only the first level should be queried and displayed firsthand and if the user expands a node the application should query again for the next level and so on.

Anyway what's the use of storing the contents of an ever changing file system in a database??