Speaking of improving and stabilizing - there is one thing which bothers me for quite a while and it is the sampling of lights and GI which they produce when those are put behind the glass both solid (refractive) and thin alike.
Recently I've been struggling with one scene which had fully modelled interior with furniture lights etc. + sun and sky + some lighmix combinations to represent the different lighting conditions - well the usual stuff. At some point I needed to do some exterior shots of the place and I had run into tremendous amounts of noise which was very persistent in the interior visible through the windows when it was rendered from outside and the lighmix pass combo's were set to represent either night or evening lights. I did some testing cause I've got a hunch that it has something to do with the way corona's gi sorts the priority\importance of lights based on their brightness and so sun and sky being around 100% more bright than any physically correctly setup interior light source takes away all the samples\priority\whatever leaving other light sources only tiny leftovers of that. Guess what, it seems, that was the right hunch, and after lowering both sun and sky intensities to 0.01 I've been able to get much cleaner interior in MUCH less time - here's some examples of the similar situation reproduced in a clean scene and hopefully clean conditions :
https://www.flickr.com/photos/119850875@N05/30780770972/in/datetaken-public/and the same both renders with lighmix set up to represent the day with the interior lights off
https://www.flickr.com/photos/119850875@N05/30261488543/in/datetaken-public/Anyone wanting to check more of those tests and different passes saved out of those is welcome to explore this -
https://www.flickr.com/photos/119850875@N05/albums/72157676287977736 album - I've done a couple of tests with different glass settings and without the glass at all, adaptivity on\off etc. the naming of files should be self evident I hope. Pay close attention to the sampling focus passes (SF) and how they differ across the different setups.
I'm almost sure it's not the best way of dealing with such conditions, physical correctness, energy conservations etc. wise, and of course it would be better if corona's GI system could\would be tweaked internally to alleviate the need of such stupid methods but it still helps me a lot to get those renders cleaner and much faster, and I'm most sure it will benefit many people who struggle with such setups as well, mainly because that nasty noise will not go away even if those images will be rendered for an hour or so, which I tried of course even if not in this particular test scene.
I wasn't sure how and where to put this report as this, most probably, is not a bug at all but is something more like a trick or just some needed internal tweak + I'm not completely sure (and who can be))it's not something wrong on my end thus it might be best if someone could reproduce such behavior too and report it too. And there's another thing inside those tests which is ligh source's disk shape being invisible behind refractive glass and visible when thin(non refracting) glass is used but that's not too big of a deal cause no one usually leaves those exposed such as here, yet still there it is ... So hoping I didn't make a fool out of myself again I post my findings here ))
Actually, It is quite easy to reproduce and everyone can test it for themselves but I'm still attaching that test scene (max 2017 and 2015) with both default and lowered sun and sky values + those lighmix pass presets for each of the scenes - no textures were harmed in the making of that test scene and the only ies file used in it is the default one which comes with 3ds max and was there for ages so hopefully it will be picked up automatically.