I think you have hit upon part of the problem: a variety of decks not only show gradual variations, but also become problematic in terms of classification.
In part it also depends on what is, for us, going to be accepted as a Tarot de Marseille. As an obvious example, its design or production in Marseille is not a requirement - rather it is a
type or, perhaps more precisely, a pattern.
Depending on how much we are going to precisely define its characteristics, we are either going to cast too narrow, or too broad, a net. I would suggest that some amongst us may perhaps want to even cast
different nets for different periods of time: a broader one prior to 1760, and a finer or narrower one post-1760.
For example, the Schaffhouse has, for me, certain characteristics that make it different from a standard Marseille-pattern: for one, XII has both feet tied. Yet I would view a very early deck depicting such but that has in nearly all other respects similar representation to the Marseille something that I would more likely refer to as 'Marseille' than 'Schaffhouse' - in part because of time.
On the whole, however, I think it is usually easier to use a concept introduced by Saul Kripke in
Naming and Necessity - that of a
RIGID DESIGNATOR. In some ways, this is a simple concept that has important ramifications: I can point to something and say "
THIS is what [given in our case our topic] is a Marseille-type deck".
And for this I would point to two early decks, and one recent one, as three different decks to which we all would (I hope) be in agreement as to their appelation as 'Marseille-type':
the Jean Payen, Jean-Pierre Payen and Jean Dodal decks (very very similar);
the Chosson and Conver decks (again, very very similar); and
the Marteau-Grimaud deck (based on the Conver in any case).
This does not mean that there are no other Marseille decks, but rather that variations will need to be considered and, if sufficiently distinctive and forming their own
GROUP of decks, perhaps another appelation given (as for the Besançon and Schaffhouse, for example).
Pointing to the Dodal and Conver of course then leads us backwards in time to consider what models
they may have used, and how or why there may have been variations: here considerations of decks such as the Vieville, the Parisian anonymous deck, the 'Charles VI', the Cary Sheet, the card remnants found in the well of one of the Visconti castles, all have a bearing on reflections and considerations.
How much deviation is, notheless, going to be 'allowed' for a deck to still be called a 'Marseille'?
Personally, I would suggest that it depends on the variation(s), and the likely
reasons for the differences.
For example, if someone copied the Grimaud but instead of the Grimaud swords used the Bolognese form (from which the Noblet and Payen, and therefore Dodal, Conver and Marteau are likely to have derived) - see the thread
10 Swords - Contrasting the Dodal and Conver - I would personally still be entirely content in calling it a Marseille.
In fact, I would personally claim it as a Marseille that has its swords historically properly depicted... even if others would see in this no more than a blending of the Marseille and the Bologna decks.
And this of course is simply an example as to what makes categorisation also difficult. Not just the deviations from the pointed-to deck (Dodal and Conver - or TdM
I and TdM
II), but what is going to be considered as intrinsically characteristic of the pattern: do the swords need to be depicted without hilts when curved? Would another name to 'Marseille' be better used in that case?
Here we may differ... and in our difference also point to not only what we value, but the manner in which we prefer various degrees of precision in our employed terms.