The way to run calibre networked is not by networking the filesystem. Frankly, the whole idea of networked filesystems is a god-awful kludge. Filesystems have the assumption of reliable, fast, synchronous I/O baked into their semantics in too many places.
Your options are:
1) Use a remote desktop solution
2) Use the content server for read only access
3) use some kind of file sync like unison or dropbox and ensure you only run one copy of calibre at a time.
4) Help with implementing the calibre networked db backend
No matter how much optimization you do, you will never get acceptable performance going through a networked filesystem. But hey, if you feel differently, calibre is open source, patches are welcome
