>>I don't know why it would show up after so long, but could it be that
>>buffering/indexing problem I've run into before, where an appended record
>>locks part of the index, and doesn't release it until the record pointer is
>>explicitly moved? (GOTO RECNO())
>
>Do you mean you always do GOTO RECNO() after an INSERT INTO?

Not always, but frequently. It was an interaction between buffering and insertion and parent-child relationships that was very reproducible, but I couldn't tell you exactly what it was.