Some twenty years ago Moshe Assis published a fragment of Yerushalmi Sanhedrin, containing most of chapters 5 and 6. Without going into detail, it is worthwhile to quote Assis' characterization of the fragment:
When we come to determine the relationship of [this] fragment with the Leiden manuscript, we discover an astonishing similarity. We find in both the same scribal errors, the same [attempted] "emendations," the same indecipherable expressions, the same inexplicable gaps.Thus, in the end the two of them must descend from the same early Urtext, together with its errors and gaps.28 Assis, later notes that the scribal errors, and erroneous at-tempted "emendations," the indecipherable expressions - all may be attributed to scribal error. But what are we to make of the gaps? He discounts the possibility that they are the result of negligent redaction, even under emergency conditions. He concludes that:
at a very early time, in Eretz Israel, the source copy, or one of the few source copies of the Yerushalmi, was damaged (nifgam) ... because of the wearing out (be-liyah) or decay (riqavon) of the parchment, or because of fading (dehiyah) or defacing (tishtush) of the ink, in [particular] places. And when the earliest scribes came to copy their Talmuds from this text, they left a blank space to indicate the place of the missing sentences. However, these blank spaces became shorter and shorter in later copies, until they disappeared altogether. And it is possible that they were already intentionally ignored at the end of the this earliest period.
As the research of Shamma Friedman over the last decade has shown, the text of the Bavli has, in many cases, been transmitted in two versions: one fuller and the other more sparse, or, most probably, closer to the original. In some cases, these two versions represent ancient traditions. Both Friedman, in his monumental study of chapter 9 of Bavli Bava Met-zia, and Adiel Schremer, in his study of the first chapter of Mo'ed Qatan, have found cases in which each branch has a different attribution for a memra and scribal error does not seem to be a factor. That both of these versions may be relatively early has been demonstrated by Moshe Benovitz in a study of the two versions of Shevu'ot chapter 3, where the fuller - the so-called "ita" version was before Rabbenu Hananel.32a This is particularly significant since the most detailed study of these two versions, that of Friedman in the intro-duction to his Talmud Arukh, employs two sources for each version: the Munich and Florence manuscripts for the ita version, and MS Hamburg and Genizah fragments for the less developed leta version. MS Florence 8 of B.M. and MS Hamburg 165 of the Bavas date to the twelfth century, the latter to 1184 and Munich 95 to 1342. All therefore date later than at least some of the Genizah fragments. And yet, while the Genizah fragments validate the authenticity of Hamburg's readings, let us not forget that one of our earliest full manu-scripts, in this case Hamburg 165, does represent the leta tradition fairly well.
Let us not forget that here too we encounter the same situation as with the Yerushalmi and the Mekhilta. At some point in the transmission of the Bavli, when few manuscripts circulated, a change in one was likely, in the course of time, to influence many more copies and copyists. At every branch in the early stemma of the Bavli, as with the Yerushalmi and the Mekhilta, one scribe, whose manuscript served as an archetype for his successors, could determine the text of the Bavli for a millennium and more. In our case, T-S F2 (2) 36  77 gives evidence of a time when our current version was only a marginal comment to that scribe, and the reading of his archetype was destined to disappear completely. But given the rule of small scale, it is probable that his text was the original; we must consider that the marginal comment was the work of one tradent, and began with one copy. But even if this was not the case, how many copies of this more difficult text circulated at a time when written texts were relatively rare?