First of all, although sometimes ellipsis is used to describe “conjunction reduction” I don’t agree with the analysis that describes coordination of lower-level constituents as ellipsis. Consider these examples:
1 a. [I want a dog] but [I don’t want a cat]
b. I [want a dog] but [don’t want a cat]
c. I want [a dog] but [not a cat]
In English, you can use the coordinators but and and not only to coordinate whole clauses, as in 1a, but also VPs (verb phrases), as in 1b and NPs (noun phrases) as in 1c. I don’t really think of coordinating smaller constituents as ellipsis for coordination of larger constituents. Rather, I prefer to reserve the term ellipsis for cases when something is actually being left out, syntactically speaking, as in these examples:
2 a. John can play the guitar; Mary can __, too.
b. Jessica had five dollars; Monica __, three.
Now, on to your actual question, which is about is whether you can coordinate unlike constituents, in particular whether you can coordinate complements of who when the complements are both object and subject. Coordination of unlike constituents is known as syllepsis, and happens when the coordinated constituents are not parallel in meaning or in grammar. Some examples of where the constituents are semantically unlike, from Wikipedia:
3 a. He carried [a strobe light] and [the responsibility for the lives of his men].
b. She lowered her standards by raising [her glass], /
[Her courage], [her eyes] and [his hopes].
c. Mr. Jones took [his coat] and [his leave].
Some examples of where the constituents are grammatically unlike, from Language Log:
4 a. CIA interrogators [stripped naked] and [played earsplitting music to] Abu Zubaydah, the first henchman of Osama bin Laden captured by US after the September 11 attacks there five years ago, according to a media report.
b. She lets the "pops" freeze partway before placing a Popsicle stick in the middle, and freezes them till [firm] and [someone wants a quick snack].
c. She wants [an engagement ring] and [her boyfriend to stop dragging his feet].
d. I called [her a whore] and [myself a cab].
Here’s a delightful example of intentional use of syllepsis from Joel Stickley,
Joe Stockley was in an expensive sports car and deep trouble. This time, he had really let his mouth and his exotic foreign lover run away with him and it was getting beyond a joke and his immediate circle of friends in the form of rumours and speculation.
As he ran a red light, the conversation back in his mind and away from his troubles, he couldn’t help but feel a sense of rising panic and the soft matte finish of his hand-stitched leather steering wheel. Angelica had been absolutely right and his wife for fifteen years, so why was he running scared, these kind of risks and this deadly gauntlet of illicit entanglements?
It’s weird, to be sure: it’s the kind of thing that spurs linguists to write multiple essays about it. I haven’t seen discussion of examples like the one you cite in Jeff Atwood’s writing, where the conjunction is both and object and a subject with who, but when I read it, it seemed slightly awkward though not entirely ungrammatical. I think part of the reason most people don’t use whom is because they have trouble identifying when it is needed, and this is because people don’t find a useful grammatical distinction between subordinate clauses that are objects or subjects. A logical result of this loss of meaningful distinction would be a concomitant loss in noticing coordination of grammatically unlike constituents with who. That is, since people don’t notice or care that there is a different syntactic form in who I didn’t know and who had built amazing stuff for us without even being asked to, then they aren’t likely to notice or care that coordinating them might be syntactically questionable.
I think most people would agree that the “surprising” effect that syllepsis gives is not appropriate in most formal writing because it draws attention away from the content of the writing and towards the cleverness of the writer.
As for the history of syllepsis with who, searching for that would be quite tedious using COHA because it doesn’t have the ability to do searches for complex syntactic structures. I’m sure that it’s not very common even today but was completely impossible before who became synonymous with whom for many speakers.