>Bingo ... Thanks to Thomas Ganss.

Glad for you.
Could you post further stats ?
This might give MS an idea to implement some kind
of include-caching, if not too difficult to realize.

"My" optimizations were done when I thought of posting
the actual numbers - so I gave only roundabout numbers.
(I was happy enough to have analyzed correctly and also
to find a working solution, since it was paid time <bg>)

Having explicit description of machines tested pre/post
optimizing should wheigh more.

regards

tg