Forked off issue833.
Could I get an investigation and explanation for why this is the case? I
imagine it has something to do with conflict marking being done on a
primitives level, but why do the sides get flipped around?
Might be an easier milestone than issue833
Indeed this needs a concrete counter-example. I have observed that the
order is different between patch formats, but since these are
completely different implementations this is not too surprising and
not a problem in practice. I am minded to close this as "works-for-me"
(but we don't have a status for that).