Data Corruption Troubleshooting FMA -- MIDRANGE-L

I have a wee problem in version V405CD mixed mode BPCS/400 & before I program
this thing I was wondering if there is a better way to do it.  Perhaps some
other BPCS site has already seen this phenomena & knows exactly what causes
it.

We have non-indexed BPCS file FMA that lists what raw material inventory is
needed for each factory order, and all programs access this data by keyed
logicals against fields in different sequences.

In theory for each order # there should be sequence # 1 2 3 4 5 etc. for each
of the different sub-component items involved.
but intermittently some orders are getting doubled up 1 1 2 2 3 3 4 4 5 5
etc. or triple 1 1 1 2 2 2 3 3 3 4 4 4 etc. same item, only some of the item
info duplicated (other fields zero in the duplicate records)

First, I want to identify which orders are corrupted so that we can kill them
& reissue them, or see if DFU delete excess is enough to uncorrupt them.
Right now the end users stumble over corrupted orders & we never know how
many other ones are out there that way.

Second, I want to identify which processes by which users lead to this
happening, since we do not know if it is a bug, improper responses to
abnormal termination, or human error.  The hope being that it is practical to
put a stop to it.  We have been using BPCS for almost 15 years & until
recently this has been an occasional glitch easily solved, but our hit rate
is on the rise, so management wants me to take the time to put a stop to it.

I figured we could use my proposed list corrupted records intermittently
between various standard user factory order management efforts to see at what
point this problem pops into existance, or the volume out there increases.

I recently learning some stuff about check constraints & wonder if there can
be a business rule that says this is a no-no, then list all no-nos.

I could write RPG program to read entire file through the old RPG cycle
looking for this scenario.
Make Order # control break L2 & Seq # L1
If I have NL2 NL1 that's a failure.

I do not see how to do this with query/400, except I wondering if I created a
logical that says no duplicate keys allowed (in the logical) then attempt
no-match physical file to its new logical to list entries in physical not in
logical.  I like to do things in query/400 that non-programmer co-workers can
then use as a guide to creating new stuff based on my showing the way.

Our disk space is a bit too tight to be adding much more stuff, so I
reluctant to explore trigger programs & journaling that would analyse every
update to locate which ones are adding the garbage, but eventually I may need
to learn how to do that.

MacWheel99@aol.com (Alister Wm Macintyre) (Al Mac)