Which aren’t? While most sensors are more invariant beyond the dual-gain threshold overall, I would still consider them “largely” invariant (at least the ones I’m familiar with). If you require maximum DR, using an ISO setting below the dual-gain point in very low light can still be of great benefit.
they are either invariant or not - and in practical terms dual gain sensors sufficiently not invariant - unless you start to narrow things down to : oh,yes - between these 2 nominal settings ISO1600 and ISO3200 they are, thus excluding away parts that not suit you ... come on
you check BClaff site for a start (unless you can DIY the same) and then you test yourself though raw conversion - because graphs might still lie to you apparently ( see @IliahBorg about R5)
Set it at base ISO with an ETTR exposure if you can.
If you want to lower read noise, have a brighter finder image, or avoid extreme pushes in your raw developer, raise it a bit, but don't get greedy and run the risk of highlight clipping. The details depend on your camera.
in an ideal scenario, when I have time/etc, I will decide AFTER ... but in a practical case I do decide BEFORE ( because I might not have time to decide precisely - so for example I select a certain Auto ISO before I start using M mode ) ... ISO not part of exposure ( process, from when exposure of a photo-site/sensel starts till that exposure ends ), but knowing how it affects the raw files (and raw processing using your specific raw converter and not abstract ideal something) becoming a part of your decision about exposure parameters to dial in ( aperture, exposure time and if you control the light, like strobes/etc - that too ) ... in that sense "ISO" still hangs around "exposure"
I don't' use auto ISO, but in the past seven or eight years I have not been in a situation where I picked my exposure based on the ISO setting I wanted to use.
I use Auto ISO and I select it once before moving to shot in M mode during some party time where I am also a part of that fun ( because I rather not fond of selecting like 2nd base ISO and pushing in raw converter or something along the lines ) - in that sense I am picking my exposure AFTER I decided about a nominal ISO ( that is I set limits for a camera to assist me and decide about the exact nominal ISO value within said limits understanding that it might not be all ideal and yes I might adjust some exposure parameters - but not auto ISO limits because that is way suits me - with what I like to do later w/ raw files in raw conversion ) - does not mean "based" on it the sense I feel you want to put in it
and if I move to use a strobe in middle because my friends want to have some posed photos under a good light then I do decide to use either a 1st base ISO or 2nd base ISO and that's it.... I will change aperture ( to get more or less people/faces/etc in DOF ), but not ISO - that I decide before based on the idea how much illumination I can deliver, etc ... so I do not change a nominal ISO after adjusting aperture/power output etc ... it is practical situation... does not matter that I agree w/ all ISO not part of exposure - it is just the way it goes in the real life
who said they havnt ? all cameras histograms are calibrated on the safe side of clipping thats all. plug a field monitor into your camera and see what a very detailed histogram looks like compared to say raw digger. on my sony a74 you can calibrate the cameras metering system and also the histogram/blinkies clipping warnings. been there done that and reset the camera back to default settings for 1/3 stop head room.
Let me go on record as saying that the statements above don't reflect any of what I have learned in more than a decade of serious, quantitative camera testing. This continued tooting of the horn of misinformation is tiresome.
Modern cameras are in some ways so forgiving that you can do things that are far from optimal and still get good images.