I don't know what the team's answer will be, but I think the answer is "they cannot offer any guarantee for any system they do not control". NFS 4 may have a standard that requires compliance with correctly implemented lock implementations, but there is nothing stopping some team from writing an implementation that cuts corners to get ahead of the competition. The POSIX standard requires that the malloc function returns null when an allocation cannot return a pointer to a block of the given size, leaving programs to believe they have all the resources they need. Linux by default ignores that requirement. If an NFS 4 implementation violates the requirement, or subsequently breaks the requirement, there isn't really anything the SQLite team can do to help provide a guarantee.