"say a reflective metallic ball is blurring to an amount X while a mirror (which is reflective metallic too, after all) is blurring to an amount Y. Then in real life you'll get a blurred reflection of a blurring ball, the blurring adds up (X+Y or something alike). But in Poser the ball in the mirror will be blurred to an amount Y only, as the blurring by the ball itself is not showing directly to the camera (but via reflection instead) and hence will not be performed in the render."

Sorry but this is not true. (It used to be how Poser worked, but not anymore.)

Here a pair of pawns - left has no blur, right has blur.

Then look behind - we see them in a blurred mirror.

It is apparent that the reflections of the two pawns are not identical. The reflected blurred pawn is more blurry than the reflected unblurred pawn. Therefore the X+Y blurring actually does happen, contrary to what you said.

Also, I did not say "As BB says, Poser blurring is a somewhat post effect." I said it's fake. Poser's focal blur is accomplished by collecting and averaging multiple samples for every pixel, but moving the objects around randomly for each sampling. This is also how it does motion blur. The two forms of blur differ only in the movement, and neither has anything at all to do with post processing. It happens during the 3D rendering. The focal blur is accomplished by having the random movement in proportion to the distance from the in-focus plane, whereas the motion blur is accomplished by having the random movement fall somewhere within the indicated motion of the object in question. Poser's motion blur movement is defined by calculating the appropriate "circle of confusion" and then moving micro-polygons randomly within that circle. This is also why Poser's bokeh doesn't look like real lens bokeh - there is no aperture involved with a shape to it.

A physical blur would not be based on moving micro-polygons around randomly - it would happen by following many sampling rays through different paths of the optical lens system. 

Poser's reflection blur has nothing at all to do with those other blurs, but it is done in a similar way - by perturbing the normal randomly (instead of positions) around the true normal, and taking lots of samples and averaging them. Originally the blurred reflection of a blurred reflection was flawed because whoever built it forgot to perturb the normal on those secondary reflections. But that has been fixed as is apparent in my render above.

However, this is an area where Poser still needs improvement. For very large amounts of reflection blur, there is always some ugliness due to insufficient samples. The "quality" parameter can be set higher than 1, but values above 1 are implemented the same as 1. This is a serious flaw, IMO. It's not a flaw of the technique - it's just a stupid programming flaw that the developer thought he knows better than us how many samples we should use.

---

Renderosity forum reply notifications are wonky. If I read a follow-up in a thread, but I don't myself reply, then notifications no longer happen AT ALL on that thread. So if I seem to be ignoring a question, that's why. (Updated September 23, 2019)