This is like #1590 but for run-in instead of tables. CSS Display says

an anonymous block box is generated around the run-in sequence

But I guess this block box should be an inline-level block container if the parent inlinifies, e.g. inside internal ruby boxes or inside another run-in with a flow inner display type. Maybe it would be better to be consistent with table fixup and also generate an inline-block inside non-run-in inline boxes.

Alternatively, in these cases the wrapper could be an inline box, or generate no wrapper.