Oh… 
No - I meant it literally! (I have a neighbour with an annoying dog!) It was a joke, but seems to have been lost in translation…
Humour is always so risky on the internet!!! 
Sorry if it’s caused you any trouble - I never thought it might be misunderstood.
This thread is bit old, but I noticed same banding issues with and other issues about dulling engine.
From my test when using dulling we get perfect average of underlying color (without real blur radius control) no matter the ‘Smudge Strength’ parameter value (which only affects opacity of blur effect)
‘Smudge Radius’ parameter is actually ‘sampling radius’ which defines how big area around brush we are sampling, and the sampling method seems to be the source of aliasing/banding problem (well it always is, even in rendering).
Here are 2 lines + 2 ellipses painted black with guides, then I draw over them with dulling brush:
You can see banding on diagonal parts. This happens if sampling radius (aka ‘smudge radius’) is bigger than brush radius. I’m not sure how other painting apps are avoiding that issue (well always picking more samples should help, at the cost of performance).
Another problem- dulling brushes are getting transparent with big sampling radius. Imo it just matter of changing order of operations - blur canvas first, then apply color to it (now it seems to work in opposite order - apply paint, blur it with background => we get mostly background color for big sampling radii).
I would also change the ‘Smudge Strength’ works. Now at 0 value it lerps dulling to transparency, but we have opacity parameter for brush transparency control. Imo Smudge Strength could be used to
lerp( pixel_brush, dulling, s_strength) #devs will understand 
I know that in Gimp you can reduce banding by swithching the image depth to 16 bit or higher. Maybe it’s worth a try? (of course you have to do the smudging again after the change)
It did not help, and would be surprised if it did, assuming that the artifacts are caused by poor canvas sampling. Anyway for future references: I figured out, that up to some threshold value of ‘smudge radius’ (aka sampling radius) the dulling works ok:
When brush preview gives uniform gray color, we get ok-ish result, and above some threshold banding artifacts are introduced. Just increase ‘smudge radius’, until the preview gives uniform gray color.
In theory when we increase sampling radius above this treshold, it should always give us perfect average gray color, because we are just averaging bigger and bigger area around brush.
I did wonder if having multiple sample points would give better results - but as you say; it would likely impact performance.
Your experiments look interesting - have you tried using flat tips? I’ve found the banding to be more pronounced when using flats, which I use more frequently than round brushes.
As it is, I’ve not been using dulling as much recently - partly because of the banding, but also because now I understand more about how it works I’m worried it might lead to an unintentional loss of saturation as I’m painting. Smearing can be a bit unwieldy and kind of like pushing liquid paint around a sheet of glass in it’s basic state; it can be tamed though and I’ve found more use for it over the past few months.
While I’m here I might as well re-state my wish to have colour smudge functionality combined with the clone engine - perhaps if I mention it enough times it’ll stick like a hypnotic suggestion and someone will feel compelled to make it a reality! 
Bump because it’s still an issue in 5.0
Are the presets you’re using set to use the new smudge algorithm (in smudge length section)?
In my testing with the new algorithm I found it to be much improved, but I think it’s still possible to get banding. If you’re using dulling presets, try changing the smudge radius as I think that can affect this.
It’s okay-ish if I use a round brush. But for flat brush tip the banding is still unacceptable. It’s sad because my ideal brush of line art is flat tip + subtle smudge.
I just want to add a little bit of smudge to Krita’s built-in chisel marker. It sounds so simple right. I can’t believe it’s still impossible in Krita.
I found a very weird non-solution tho… for some reason the banding (?) is more obvious at the start of a stroke.
So I set a fade-in color rate to counter it:
It’s not a solution, not even a good workaround either. It’s just something I’m using.
If you can come up with the mathematics that would make it possible, it might be possible to code, but I suspect it just isn’t that simple.
Yes, I guess so. I shouldn’t have implied it’s simple. It’s just that Krita’s huge amount of brush options gives me illusion that it can simulate other apps with simpler brush engine (like Sai). But of course it’s just my illusion.
That being said, this problem is so much more obvious with flat brush tip than with round brush tip. That makes me think it’s that “Anisotropic spacing” is not correctly implemented. Or it’s technically correct, but it’s not enough for color smudge brush. We might need to make the spacing artificially denser (at the cost of performance) for color smudge. It’s just my wild guess, tho.
Another possible solution is to stretch smudge radius along the brush’s direction?
