Jim,

> No, happily this isn't possible. Blobs are garbage collected when the
> parent records are deleted or have versions pruned. And the symptom
> would only be visible from a read committed transaction and then only
> between the interval of fetching a record and having it deleted and
> garbage collected. A transaction containing an insert wouldn't meet any
> of these conditions, let alone all of them.

Even if we talk about "INSERT INTO ... SELECT * FROM aTable"? So, if we have
READ COMMITTED isolation and engine fetches the record including the BLOB ID
and before the record is written, parent is deleted and BLOB is garbage
collected. Can this happen?

Roman