NW6.0-SP4 with BMEE3.6

Health-monitor reports server running low on available
directory entries. There are now around 3% left.

First thought was to increase Maximum Directory Cache
Buffers. That somehow did not work. Guess I need to add it
to startup.ncf and reboot the box, what I do not want right

Since the bm-proxy-cache-volume (traditional) is filled with
around 55.000 files, my second idea was to clean out this by
unloading BM and deleting the cache-directories.

At least both paths seem not to be the normal way how to
handle this.

Any ideas are highly welcome.