I have now claimed a few times on the internet, based on something (sensible!) I read, that at some point in the 1920s, that Zermelo at one point considered as a set theoretic axiom (schema) something equivalent to a restricted version of Replacement, corresponding to a function with countable domain, before going all out by 1930. Now, however, I cannot find the document in which I read this, and I cannot remember anything else about it! I foolishly didn't write this down. I have consulted Ebbinghaus' book, and I can't see it in there. I have consulted Kanamori's edition of Zermelo's collected works, and I can't see it in the prefatory essays on Zermelo's 1929 or 1930 set theory papers.
However, and this is where my faulty memory shows up even more: I also in one place claimed it was Fraenkel who considered countable Replacement! I cannot find this either. And in trying to hunt down my original source, I recognised the phrase
it is also to be noted that Skolem only wrote that "we could introduce" Replacement
in Kanamori's 2012 In Praise of Replacement. But I didn't see, on a search through there, anything corresponding to my half-remembered claim.
Can anyone enlighten me?