Cache on demand: how to define the extent?

[MarkusSchenardi]

I'm running a webapplication with cache on demand set for detailed scales (1:100, 1:250).

When calling a not yet existing cache image, the server is calculating the images around my desired extent with a buffer of at least 7 times the image size.These way it takes to much time until the really desired tiles get displayed.

Question: Is it parcgisossible to adjust the buffer size to safe server resources and to improve the time until the desired extent is being displayed?

AND: why are cache on demand tiles processed on the service-SOC itself and not with the Caching-Service (GP-Service)?