We have cases where we have a parent, child, grandchild, great-grandchild data relationship (batch, invoice, lines, detail account transactions).

We typically add one batch per user per month, then all the invoice records are added to that batch for that time for that user. Each invoice is a thing (not just invoices, but POs, checks, etc.), with the lines being line item detail, and then the great-grandchildren are the accounting transactions related to those line items.

We're seeing cases where (as these records are written in rapid succession when a new line is added) the line is recorded, and sometimes it will not record the detail account transaction records. Or it will record some of them, but not all. This usually coincides with a corruption in the index that then causes other data issues after, but our customers usually find it out immediately because their input screen has become affected.

This particular type of recording everything except the detail lines has now impacted a small percentage of our customers, and we can't find any pattern or reason for it. I've actually created a generic fix program that our support department can use to correct this issue without us having to investigate every single one. We're also trying to have our support department query the customers to see if they can remember anything unusual they did when it happened.

Our system records various logs indicating data changes, events, times, etc., and we can sometimes determine that the issue is caused when two people record something in rapid succession and it's an SMB issue. But we've had our customers alter their cache settings so those are not typically issues any longer. Still, it's a new thing that's just started happening in the past few weeks.

We've more or less concluded it's something happening due to a Windows update, but it doesn't affect every computer or server, so there's something somewhere that's not like the way it is in other drivers, or on other hardware. But, we just don't know.

>Interesting... what kind of data failure?
>
>>We've seen a sudden an unexpected increase in a particular type of data failure. Not every user is seeing it. Only about 5%.