But it does really only show at high iso, and only in somewhat specific cases. Even at ISO 1600 its usually not an issue at all, and it will sure as hell do a lot better then a 20/4.0 (which is what most alternatives at that price range will give at 20mm) needing ISO 8000 approx with the same amount of light.

Does that mean the issue doesn’t exist? No, not at all, but it does mean you are only going to see this issue in rather specific cases.

If you were looking for perfection, you can stop right here and now, only theoretical lenses can be perfect, every practical lens is a compromise and not perfect by definition. All of them have situations in which they don’t perform too well or even ‘misbehave’. Yes that includes nice lenses like the Pana/Leica 25/1.4, or that $1000000+ spy glass the NSA and alike use.

So decide for yourself, based on actually inspecting results instead of exaggerated interpretation of already exaggerated statements from random anonymous people in some internet forum, if the results and price are good for what you need, or if you should be looking for something else.

My point was: IF image quality of 20/1.7 and 17/1.8 are about the same and IF you own an OM-D E-M5, then the 20/1.7 might be worse choice. I just wanted to mention this. I was surprised that this wasn’t pointed out in this thread yet. Again: Saying “The 20/1.7 is the much better choice because it is much cheapter” only is half of the truth.