As far as I can tell from reading https://drafts.csswg.org/css2/visuren.html#z-index only position fixed implicitly creates a stacking context. Loading http://output.jsbin.com/yucorey shows an example page which demonstrates that currently sticky is creating a stacking context (i.e. the z-index 2 descendant is stacked below the z-index 1 sibling.
I think it's better to match position:fixed here, since sticky has fixed-like behavior part of the time. https://drafts.csswg.org/css-position-3/#position-property needs to be updated to say this.
https://github.com/w3c/csswg-drafts/issues/1053
@smfr: Did safari regress something here? data:text/html,<meta charset=utf8><h1 style="position: sticky; top: 0;background-color:tomato; text-align: center;">Hello</h1><main style="height: 200vh;"><button>Plain, under</button><button style="opacity: 0.6;">Translucid, overlaps</button></main> Seems to not behave the same as: data:text/html,<meta charset=utf8><h1 style="position: fixed; width: 100% top: 0;background-color:tomato; text-align: center;">Hello</h1><main style="height: 200vh;"><button>Plain, under</button><button style="opacity: 0.6;">Translucid, overlaps</button></main> in Safari, but it does in Gecko and Blink.
Looks like we have an overlap-testing bug where "Translucid" fails to z-order on top o the sticky if you scroll (it gets fixed if you resize the window when it's on top).