Ask HN: Intentionally inject errors into NFS to test user client

We have a software client that is being deployed into different IT environments. This client reads and writes files and thus has to handle a variety of different filesystem/networking conditions. Most customers' filesystems are mounted via NFS and a lot of times we encounter problems reading and/or writing files in certain directories because of NFS delays and improperly configured NFS permissions.

I'm looking for ways to intentionally inject errors (NFS/networking/etc.) in our test environment so we can better handle such conditions.

Searching around, I found mentions of a Linux kernel configuration item called CONFIG_NFSD_FAULT_INJECTION that supposedly allows "manual injection of faults" but I've been unable to find a working example so far.

Fellow HN-ers, what have you done or what would you do in such a scenario? Any advice is much appreciated!