High volume of File IO and the Network Redirector

I am observing some strange behavior from somewhere within the network redirector (or server).

I have a service that runs as SYSTEM and and mini filter driver. For some strange reason when my driver is running, GetFileAttributes() sometimes fails with 0xffffffff for a file on a network. Using process monitor I captured the API traces.

There are ~800 success calls to share XYZ (within 1 second). Then there are ~800 failed calls to the same share (within 1 second) - the first few calls fail with STATUS_ACCESS_DENIED (which are expected - I don’t impersonate) and the rest fail with STATUS_BAD_NETWORK_PATH. Then it starts working again, another 200 or so success calls.

The GetFileAttributes executes during the 800 failed operations from another process running as the Logged on user and for some reason it returns STATUS_BAD_NETWORK_PATH. But this works fine if it happens that this runs before or after the 800 failed called by my service.

I modified my driver so that it issues less File IO and the problem is solved. From 800 failed, there are 250 failed now. But GetFileAttributes() now works always! This is very strange.

Is there anything or some kind of threshold that I am hitting that leads to this behavior? (Denial or service attach prevention mechanism or some kind of Quality of service)