Hidden render elements for denoising explanation

[enthanion]

Hello!
I'm trying to understand what additional data is stored inside the CXR files. There is some info about the format in wiki (https://corona-renderer.com/wiki/corona_exr) but not much about included channels (the only thing mentioned is that they are here for denoising). Here's what I can guess, correct me if I'm wrong:
VisibleDiffuse - CShading_SourceColor
VisibleNormals - CGeometry_NormalsShading
Hybrid - Is this some kind of ZDepth? What is it's range?
HairIds - ?
Prefixed by Sqr - 1.0 gamma versions of non-prefixed?

[maru]

We can probably provide explanations what those secret elements are, but why exactly do you need this info?

[enthanion]

The main reason - I want to to generate them manually. Well, if it is some sort of secret info you can't share, I'm ok with it.

[maru]

The main reason - I want to to generate them manually.

But why? :)

[enthanion]

But why? :)

Well, here's the situation: I've got a rendered frame that doesn't contain these elements but I'd like to denoise it so I need to get these elements somehow. A possible solution would be to use Resume from file... but as it is stated in this article the resulting image might not be denoised properly.
In V-Ray's denoising tool you can easily generate the missing elements if you need to so I started wondering - if it is any similar in Corona's one? As it turns out, not really - these elements are hidden in first place (you can still access them using any exr-viewer\editor) and some of them look like specific ones you can't generate using available render elements anyways. Or can you?

[maru]

Currently this can only be done with the Resume feature. If enough passes are rendered with denoising enabled, it should work fine.

I have additionally logged a feature request, but it is not going to be high priority.
(internal note id=312883432)

[enthanion]

I have additionally logged a feature request, but it is not going to be high priority.
(internal note id=312883432)

Nice to hear, thanks.
As I understand, the denoiser will only consider noise info from the passes that were done after resuming, so you lose all the info about it before, right? Is there any way to make denoiser use fixed noise threshold for the whole image rather than deriving it from passes? Like manipulating CORONA_FB_WEIGHTS/Hybrid somehow?