Just to be 100% sure, I assume you turned off AV real-time scanning
on the server computer as well, if you have one on your network.

I'm wondering if you might have a problem with a particular 
workstation - network card, bad cable etc. Is there any way you could:
- ideally, test with 2 other completely different workstations
- failing that, try to test with one other different workstation, against 
first one of your original test workstations, then against the other

Yes, I've tested different workstations/server combinations on different networks, etc. Same result.
I've tested XP to XP, Vista to Vista, XP to Vista, Vista to XP, XP to Server 2000, Vista to Server 2000, XP to server 2003, etc. They all have the same kind of performance numbers.

I'm also assuming the networking is all wired, I highly recommend 
against using a wireless connection for file-server DB applications such as VFP.

All hard-wired, no wireless.

Is there any unusual utility software running in your environment, 
that might hook into the file system? AV is an obvious example, but 
there may be others such as real-time backup programs, file system 
monitors etc. Basically, any software running on server or workstations 
that might be considered odd or unusual?

Unfortunately, nothing unusual. Just plain old computers running on plain old 100mb networks.

This may not affect you, but I got a tip the other day for general network
performance improvement. Turn off all unused protocol support on all your network devices.

That's a good tip.


There has got to be something I'm overlooking that explains why:
The same workstation accessing the same data across the network will perform this routine in 7 seconds as long as no other workstations have the database/table open. As soon as another workstation opens the database/table, even if the other workstation isn't actually doing anything, it makes the routine on the first workstation slowdown from 7 seconds to 35+ seconds.
Another note: in a peer-to-peer environment, the "server" computer that hosts the data does not slow down when another computer opens the data; the "client" computer (as mentioned above), however, does.

I'm going to try and drill down using the coverage profiler in my compiled .exe to figure out what is actually taking so long. Any tips on a good coverage profiler add-on?