This opens a can of worms. In all three cases, the reason for putting the instruction in the previous bar is because there is not enough room above the note. Despite this, it is still readable and perfectly obvious to what the instruction refers. However, modern computer music setting prefers to avoid doing this, and consequently uses more space in ensuring that nothing overlaps. Many modern editions of Classical and Romantic scores have less bars per page than older ones for this reason, and the conductor has to turn over pages more often. The Mahler scores are a good example. The latest edition of the Fifth Symphony score takes 333 pages, versus 244 of the earlier scores, and the Sixth Symphony takes 346 versus 261 pages. I personally find them less legible and prefer to use the earlier scores, which are beautifully engraved, while breaking the computer setting rules of placement of text, slurs, hairpins. Like many composers’ manuscripts, which also break the rules they are perfectly legible to a performing musician.
One might hope for an option to be established such that such preferred placements as you have mentioned are applied to both score and parts in certain cases. I add this caveat because there may on occasion actually be room in the part for the text to fit in the normal place, although there was not in the score, and one would also want on occasion to avoid the splitting of a multirest to provide an empty bar for the placement of text.
It is not a simple matter to apply artificial intelligence to music setting that has profited over the years from intelligent modifications of the rules, guided by human intervention.