>>Thanks - the problem occurred after we stopped getting the read errors due to the mapping problem. I can see where the problem I am seeing happened a few times before, but only once since changing from using the mapped drive.
>>
>>It is really strange - I can open the table, set the order to the order number, then seek the number that had the problem, and it takes me to a record that should have the order number in it, but the order number field is blank. Does anyone know what sequence FoxPro updates tables/indexes? I always assumed that it added the record to the table then updated the index. This looks like it somehow updated the index to reflect what is supposed to be in the record, but didn't get the data in the record....
>
>My understanding is the same as yours, the table should get updated before the index. If it isn't, that implies something interfering with normal file system operations.
>
>One thing that does that is antivirus programs. Maybe you could test with real-time scanning temporarily disabled on all machines that access or host the data, or exclude the VFP data and temp file folder(s) from real-time scanning.
>

It might be a side effect of buffering: the record write only adds a small amount of info to the buffer, whereas an index entry add MIGHT create a rewrite of a couple of pages while (re-)balancing the structure. Since Buffer>>File is sometimes flakey due to SMB2/Oplocks, the index changes might be earlier on the physical disk while the record write still is in the buffer.