(Under-scoring Keith's point here:) So, despite the clear and experience-based reasons that separation of the running SQLite library from its persistent storage by the interposition of network file-emulation layers is specifically recommended to be avoided, you want to use a particular variant of one of those layers which is purported to improve performance in some way? Does chachefilesd improve the odds that all readers and writers of the database "file" see/create a consistent view of it? Or is it a speed-up fixer-upper? > \[ cachefilesd does not seem to speed SQLite queries \] Does anyone has any idea why? Perhaps cache-improver is taking care to be sure that the underlying actual disk file on a remote machine has not changed underneath the cached copies. There is already a lot of caching and syncing effort associated with network file-emulation layers. \[quoted out-of-order\] > I know, I know! It is not recommended to use SQLite over NFS. I suppose that means you have read and understood [SQLite Over a Network, Caveats and Considerations](https://sqlite.org/useovernet.html) and [See How To Corrupt Your Database Files](https://sqlite.org/lockingv3.html#how_to_corrupt). Do you have any suggestions as to how such dissuasion might be made more effective?