DSXchange

Where?

Not at all. You can build your Lookup File Set on the five digit key, and allow duplicates. In the Lookup stage that uses it, allow multiple row returns from this reference input link.

You may as well use a server job - this whole thing is going to be sequential in any case.
Construct rows in a Transformer stage that contain the requisite data and newline character. Write to target sequential file as a single VarChar column with no field delimiters.

... are you the Moderator ?

I think you'll need to contact your vendor/support provider over this one.

What changes did you make in uvconfig?

Shouldn't you supply the new value with the -itag option? (That's from memory, but I recall that -itag ade is needed to reset to ade.)

Why does customer want two separate projects?

Why does the "solution" not involve scheduling the job in the two separate projects (for example from a third party scheduling tool) rather than forcing DataStage to do a job that it is not designed to do?

Error codes in the 39xxx range are related to the architecture underlying DS Objects through which the DataStage clients and their server-based agent processes communicate. The most comprehensive listing of these error codes is the Intercall Reference Guide, which can be downloaded from IBM. Note th...

The fact that you're using inter-process row buffering. A timeout is interpreted by the downstream stage to mean that something fatal has occurred upstream, so it also will abort. 100 sec is probably erring too much on the side of caution - an awful lot of rows could be processed in that time. If yo...

If you "know" that, your knowledge is five or more years out of date. It used to be UniVerse; it stopped being UniVerse with effect from version 6.0.

DSJ.ME as the job handle is expressly forbidden to DSAttachJob() function.

Start by reading the manual that covers these stage types (XMLPACK_20_Designer.pdf). The descriptions therein are quite easy to follow.

Nice logic. Would have been improved by use of the correct function names, such as MonthFromDate() and YearFromDate() in the Transformer stage. Also, ought you not to compare MonthDayFromDate() values to determine A or B approach, rather than the actual dates?

It will still be slow unless the column(s) named in the WHERE clause are indexed.