Thread by @PhDniX

One of the oldest known Quranic manuscripts, the Codex Parisino Petropolitanus, marks verse markings quite regularly. However, sometimes we end up with verse markers (six small dots) in places we don't expect them. For example here after سبيلا sabīlan in quran.com/4/34/

Déroche records several more of these in this manuscript.
quran.com/4/79/ at rasūlan (erased)
quran.com/9/115/ at mā yattaqūna (erased?)
quran.com/5/3/ at bi-l-ʾazlāmi
quran.com/10/10 at salām (erased)

quran.com/14/27 at al-ẓālimīna
Also at ʾāxarūna at quran.com/25/4 but I do not have a photo of it.

So what exactly is going on? In 'Qurans of the Umayyads' Déroche claims that these 'verse breaks' are unknown in the traditions on verse division.

These then, to him are evidence for a not quite canonized verse divisions. And he suggests that these "short verses" that are created of an editorial process of the Quranic revelation whose traces are seen in this 7th century manuscript.

That's a bold claim that does not rest on a whole lot of evidence. It is in fact wrong to say that these verse divisions are not recorded by the tradition.

These verse breaks are indeed not recorded in Spitaler's overview on verse counts, but traditional works record them.

In al-Bayān fī ʿadd ʾāy al-Qurʾān "The explanation on the counting of the verses of the Quran" by ʾAbū ʿAmr al-Dānī, we find that he consistently records verse breaks that not a single one of the regional codices counts count towards the total.

Non-counted verses are a puzzling category -- what is the point of marking a verse, calling it a verse break, but then not counting it? Yet, this is something that is recorded in the tradition which we also find in manuscripts.

Arabe 334a marks with a circle every time that it hits a 10th verse. Because of this, we can confirm that some of the verse markers found in that manuscript were not meant to be counted by the scribe himself -- yet they are marked, e.g. muʿāǧizīna in quran.com/34/38

For some of the 'short verses' found by Déroche these are -- contrary to his claim -- in fact recorded in the tradition. Just annoyingly, not in the schematic recording of Spitaler. But when secondary sources fail us we have, of course, to go back to the original sources.

Q4:34 sabīlan; Q4:79 rasūlan; Q9:115 mā yattaqūna and Q25:4 qawmun ʾāxarūna are all recorded by al-Dānī.

Three remain that are not recorded by my favourite medieval Andalusian:
Q5:3 bi-l-ʾazlāmi; Q10:10 salāmun; Q14:27 al-ẓālimūna.

If we examine these verses in a modern print Quran, we noticed they have something in common: They're all marked as optional pauses (miniature jīm)

Also the 'non-counted verse marks' recorded by al-Dānī are all marked as either being optional pauses, or slightly discouraged pauses. All of them form complete sentences. What Déroche takes as "verse marks" in these early manuscripts seem to have functioned as a semicolons.

That these function as semicolons rather than verse was already observed by @CellardEleonore in her PhD thesis.

It is very important not to exoticize these 'uncounted' verses too much. It is clear that early on, they were not considered part of the count.

With Arabe 334a it is even clear that the same hand who wrote the "verse markers"-cum-semicolons also wrote the 10-verse markers. He therefore clearly chose to not count those pseudo-verse breaks despite writing them. Clearly the 'verse marker' had more than just that function.

It is, in fact, impressive that verse counts tend to align well with regional rasm variants. For example Arabe 328 and Or. 2165 are Syrian both in rasm and verse breaks/counts.

It seems that the verse breaks and their counts were part of the local traditions very early on.