For example whats real problems this bug may create http://take.ms/isanc and i am no idea how fix it

I see no problems in https://jsfiddle.net/f4nd7tta/41/ - I see red dots to the right of every grey rectangle. Isn't that what we're supposed to see?

Note that break-inside:avoid (FYI: "column-break-inside" is not a valid property name - it's "break-inside") does *not* mean that it's *forbidden* to break inside an element. We'll still insert breaks at valid class A, B or C [1] break points to prevent column overflow.

https://jsfiddle.net/f4nd7tta/42/ has a multicol container with relatively positioned children. Each relatively positioned child has an absolutely positioned child. The containing block for those absolutely positioned objects is their relatively positioned parent, whose containing block is the multicol container. As such, the absolutely positioned objects also take part in the fragmentation context. So I don't think we're doing anything wrong here either.

[1] https://drafts.csswg.org/css-break/#possible-breaks

Strange, here https://jsfiddle.net/f4nd7tta/41/ what I see http://take.ms/ZBG1P
Maybe you watch an example on the screen with high resolution, excuse me for this mistake.

https://drafts.csswg.org/css-multicol/#overflow
"Floated or in-flow content that extends into column gaps or neighboring columns — e.g., long words or images — is not clipped and may therefore cause overflow."
however what we see in Chrome https://jsfiddle.net/9a276nbd/ http://take.ms/JNViO

And https://drafts.csswg.org/css-multicol/#pagination-and-overflow-outside-multicol
"Content and column rules that extend outside column boxes at the edges of the multi-column element are clipped according to the overflow property."
however what we see https://jsfiddle.net/mqr25637/ http://take.ms/ziU3Z
 
Why absolutely positioned objects are a part of fragmentation context in multicol element, and where I may read about? multicol container don't create new document with own flow and it is a part of a normal flow of the document as well as his childrens. And as we know absolutely positioned elements "removed from the normal flow entirely (it has no impact on later siblings)" https://www.w3.org/TR/CSS2/visuren.html#absolute-positioning It is mean absolutely positioned elements have to be removed and from multicol container, if it is so possible to tell. I consider, page fragmentation https://www.w3.org/TR/css-break-3/#transforms it is not same for absolutely positioned elements then any other fragmentations context created by line break or column break, example for line break, https://jsfiddle.net/rao3kubm/ http://take.ms/zNp2G in this example absolutely positioned element have relatively positioned parent and both of them are located inside linebox(like column box), but this absolutely positioned object isn't a part fragmentation context.

I very much you ask to look narrowly at this problem once again. Similar behavior absolutely positioned elements in multicol conteiner strongly limits use of them with css multicol in real projects.

Yeah, my browser window is 1600px wide, so I missed this. If I make the window narrower, I can indeed see that the red dots get clipped at column boundaries. This is a spec violation according to the Editor's Draft of multicol. However, Blink follows the CR version of the spec here, which says roughly the opposite: https://www.w3.org/TR/css3-multicol/#overflow-inside-multicol-elements

"Content in the normal flow that extends into column gaps (e.g., long words or images) is clipped in the middle of the column gap."

Of course, in the test, the red dot is an out-of-flow object, so one might claim that the above rule doesn't apply. Then again, the out-of-flow object is contained by an in-flow relatively positioned object, which is contained by the multicol container, so it really makes sense to me to clip it.

According to CSS 2, the containing block of an absolutely positioned object is the nearest ancestor that has a "position" different from "static", e.g. "relative". In the test, that is the parent. See https://www.w3.org/TR/CSS22/visudet.html#containing-block-details . Later CSS modules extend this somewhat, so that e.g. the nearest ancestor that establishes a transform also serves as a containing block for absolutely positioned descendants.

It is true that absolutely positioned objects should not affect siblings or ancestors (with a few exceptions), but they ARE affected (and clipped) by ancestors.
<p>There should be no red here.</p>
<div style="position:relative; overflow:hidden; width:100px; height:100px;">
    <div style="position:absolute; left:100px; width:100px; height:100px; background:red;"></div>
</div>

Thank you for the detailed answer. I apologize for the fact that has taken away from you time, today the condition of the specification are disappoints. Let's hope, that these mistakes in the specification will correct soon. 

We should consider implementing what the spec says now (don't clip anything automatically), rather than what it used to say (clip at column gaps), so marking as duplicate rather than wontfix.

https://www.w3.org/TR/css-multicol-1/#overflow-inside-multicol-elements
8.1. Overflow inside multicol elements
Except for cases where this would cause a column break, content that extends outside column boxes visibly overflows and is not clipped to the column box.