The peculiar agricultural terminology commonly used in algebraic geometry and category theory, "sheaf", "stalk", "germ", is quite well-known. A sheaf is pictured as something like a bundle of stalks, in which reside germs. Very roughly and intuitively, a germ is a localized datum capable of being developed or extended to a function.
It is commonly noted that Jean Leray introduced sheaves (and also spectral sequences) while he was interned as a prisoner of war (in Austria during World War II). In French, the terms "faisceau, fibre, germe" are used, with faisceau making its first appearance in Leray. The term "fiber" is another well-known term in mathematical English, as in "fiber bundle", etc. -- such terms are conceptual neighbors of the sheaf-theoretic terms above, although I suppose the need was felt to translate this sense of the French fibre into something other than "fiber" (for example, the fiber of a vector bundle over a point of the base space is something different from the stalk (over the same point) of the corresponding sheaf of modules). Hence the English "stalk".
I am interested in knowing more about the provenance and etymology of the mathematical terms. Here is the precise question:
Are there any sources that attest to why Leray chose faisceau? Relatedly, are there earlier sources (predating Leray) which use for example the term "germ" (or any of its cognates in other languages)?
(I must confess that I haven't looked at Leray's seminal papers. I'm pretty sure I'm not alone in this, as many mathematicians from that time commented and complained about how obscure Leray's presentations were, and everyone nowadays seems to learn sheaf theory and spectral sequences from other sources.)
I might as well share my hunch or private folk etymology which leads me to ask this question: it was "germ/germe" that came first, with other terms like "faisceau/sheaf" and "fibre/stalk" later being built around it, and that "germ" itself might be rooted in the theory of Riemann surfaces, with later carry-over into the modern foundations of differentiable manifolds and algebraic varieties. The idea is that starting with a germ of an analytic function, i.e., its local behavior in arbitrarily small neighborhoods of a point, there germinates by analytic continuation a development into a maximal connected Riemann surface in which the analytic function naturally "lives" (possibly as a multi-valued meromorphic function). I think I probably first got this idea from Chapter 8 of Complex Analysis by Ahlfors.