>>>>>Backend you were working against? .WhereType you are using ? Which PK-Type and (depending on previous answers) how created ?
>>>>
>>>>SQL /(20005 and 2014, possibly 2008r2 as well), pk only, int, created manually (i.e. calling a server-side function independently, before tableupdate).
>>>
>>>Conflict strategy / .WhereType you are using ?
>>>
>>>As this is some kind of tree/recursive table, 1st CA probably needs to tableupdate() for an error to manifest at all, but nothing written above or necessary about 2nd CA. "Same record" of 3rd CA refers to newly created record from 1st CA, last from 2nd (implying tableupdate() there, all "new" records or some greater set?
>>>
>>>We have sometimes summary CA working on R/W tables, we have recursive tables for instance for rights, often used in security joins/filters, but probably not your scenario.
>>
>>OK, more details. This is two tables with persons in them; the first one contains candidates (coming from the web, so it may contain spam). This happens when a candidate is approved, its record is inserted into the people table. The 2nd bizobject is in charge of inserting the phone into the phones table and updating a FK field in the people table, then it tableupdates both (phones and people). The 3rd bizobject then just retrieves this record as if it would retrieve any other people record, for display.
>>
>>While I was doing the safe select into this 3rd bizobject's CA (i.e. senduptates=.f., zap, select into a temp cursor, append from it, all of a sudden the .sendupdates=.t. and this append is now an insert, immediately sent to server, key constraint violation, bang. Tried to work around this in several ways, and in the end only .cursorfill with a new .selectCmd worked.
>
>As the 3rd bizobject retrieves a "record", is that from a single table or a join from people to (main or many?) phone to display both?

It has five joins to a different table (a list of lists); in detail view phones are handled separately. IOW, this third bizobject doesn't have a "f" from phones.