I'm seeing a related issue in that when I try to use an ITVF as the source data for an INSERT INTO a table (using minimally logged operations), where the ITVF is at the root of a hierarchy of nested ITVFs, the performance of the INSERT INTO totally sucks. We're talking fairly large resultsets (6-12 million records).
I'm scratching my head, and I've been at this for some time.
Any comments or thoughts.
Respectfully,
Darryll