I don't understand what the ruckus is about. Looking at the samples, simple observation yields two points:
a) the little superscript letters give an immediate and powerful "guide to the eye". There simply is no way you can be confused as to the writing direction of a text snippet (as apposed to isolated letters). Anybody who is fluent in this script will read "word shapes" as you would do in Latin, not "letters". There are apparently so many superscripts that they completely compensate for the issue of symmetric shapes for the base letters.
b) The same superscripted letters would seem to make it exceedingly awkward to write texts in vertical mode with upright glyphs. They give all appearance of forming logical clusters or variable length. Those are better accommodated in vertical flow by rotating the line as a whole (hence rotating the glyphs).
c) Even if the superscripts don't form graphical or clusters with adjacent characters, the fact is that they take so much less space. This makes it awkward in vertical "character at a time" mode, because many cells would be near-empty.
The example of Latin was given, but I think it's misapplied. In most languages, Latin characters do not form strong clusters, and where they do, that might be limited to certain type styles. Dutch IJ is sometimes written as a "unit" in signage that uses upright glyphs for vertical text. Fraktur "ck" clusters strongly enough to not acquire intercharacter spacing in justification (I've not seen any vertical + upright samples of Fraktur). But these types of letter clusters are the exception. For the Syllabics, clustering seems to rather be the rule.
Certainly, if you wanted to cite syllabics in vertically set East Asian text there would be no doubt that you'd want to rotate such text.
A./
