There's a separate-ish issue: collapsing/expanding the sidebar after switching to edit mode won't notify the toolbar that it needs to redraw, so the width winds up being incorrect. (Covering the sidebar / being misaligned.)

@ppelberg I'm not sure what the screenshots in the description are being used for but just to note: "VE toolbar wide as page" only occurs when the browser window is narrow (~1,000px or less).

in case these are helpful here are some high level options for the sticky VE toolbar/header:

1. Toolbar/header container is full-width, but toolbar is contained to article content width

wider window	narrower window

2. Toolbar/header is full-width

wider window	narrower window

3. Toolbar/header is contained to article content width
(this is what's currently happening)

wider window	narrower window

depending on what you all are thinking it could also be worth considering how the table of contents (or a different menu in the sidebar) would affect this, for example:

option 1	option 2	option 3

I think the problems with options 1/2 are what happens at the top of the page, and what happens when you transition from the top of the page to/from the middle of the page.

The specific awkward transition between read and edit at the moment, incidentally, is when you're scrolled down enough that you can't see the sidebar, and use the sticky header's edit button or a section edit link to enter VE. It results in the full-width sticky header fading out and a partial-width VE toolbar fading in to replace it, which feels awkward. (But might be unavoidable given all the possible states.)

In T299113#7690260, @DLynch wrote:

The specific awkward transition between read and edit at the moment, incidentally, is when you're scrolled down enough that you can't see the sidebar, and use the sticky header's edit button or a section edit link to enter VE. It results in the full-width sticky header fading out and a partial-width VE toolbar fading in to replace it, which feels awkward. (But might be unavoidable given all the possible states.)

this sketch, which I posted above (T299113#7672103), was intended to resolve that awkwardness to some extent: