S. Douglas Olson, "The Fragments of Aristophanes' Gerytades
Methodological Considerations," in Anna Lamari et al., edd., Fragmentation in Ancient Greek Drama
(Berlin: de Gruyter, 2020), pp. 129-144 (at 142-143):
Since assertions regarding fragments inevitably involve some degree of
guesswork, we ought to ask what sort of guesswork and in particular how many
mutually dependent hypotheses we are willing to tolerate in discussions of them.
The obvious danger is elaborate "house of cards" arguments, for the perverse nature of such constructions is that every additional "card" added to the house reduces the likelihood that the thesis as a whole is correct rather than increasing it.
If a thesis depends on two proposals, for example, each of which has a 50%
chance of being correct, the thesis as a whole has only a 25% chance of being
right, and every additional such proposal added to the structure lowers the odds
even further. "House of cards" arguments tend to be recognizable by their elaborate and ingenious character, with one arbitrary intuition or conjecture succeeding another until the grand conclusion is drawn. They also tend to be marked by
phrases such as "if we assume", on the one hand, and by the use of words such
as "clearly", which elicit the reader's assent without arguing the point, on the
other. Without offering any specific quantitative criteria (since opinions may differ on the point), I suggest that any hypothesis regarding a lost text that can reasonably be described as a "house of cards" should be rejected a priori; this is a
bad way to argue, except once again to the extent that such arguments contribute
to other — contemporary — ends that have nothing specifically to do with ancient
texts, even if ancient texts serve as a means of discussing them.
Id. (at 144):
Citation of an authority, above all else a modern authority, is not an argument. In particular, this means that further hypotheses ought not to be built on
the conjectures of others even if those conjectures have ossified into scholarly
doctrine; this is merely to assist in the construction of a multi-generational
"house of cards", which looks secure because every individual scholar adds only
one or two more hypotheses to the construction.