JCL, although this sounds quite sensible... It can not excuse the framerate produced by many entities in A6 compared to other engines. Yes, once again someone comparing engines, even at the risk of becoming very unpopular in this place here! Let's have an example. The background-level for my current game project's main menu has got a lot of grass/vegetation models. These have the alpha-test material (->wiki) applied and all of them have actions running.
When opening up the control panel via [F11], it shows me:
Script exec time: ~1ms
Entity render time: ~25ms
Ref (I believe this one shows the calc time for the alpha-test): ~0.3ms

This is the scene, by the way (please ignore the water shader, as switching it off made a difference of about only 2-3 frames):

Here you can see that most of the calculation time is consumed by the entities themselves, not by the alpha-test material or the scripts. Nevertheless I tried to explore this further on. I rewrote my scripts, so that the models don't have any action at all now, the alpha-test material is still in but does nothing but alpha-testing (e.g. no dynamic deformation or sth. similar). So what I finally got is a level with nearly all actions removed, featuring about 500 entities with alpha test and small skins (64x64 and 128x128).
Except for the script exec time that changed to ~0.2ms, nothing happended and all those measures brought a plus of one single frame. There must be something wrong there! I'm using a Geforce 4 Ti 4200, which even does the job for, say, Farcry (if you reduce some details), where many many vegetation models are in the view. And I strongly believe this has to be an A6-issue as even making all the grass models use the same texture (mtlSkin1 instead of entSkin1 in the alpha-test effect) brought nothing but one more single frame.
Correct me if I'm seriously missing something here, but this sounds very strange to me.