As you correctly observed, English is tricky and there aren’t too many hard rules with the language, which is aggravating for English learners but also highlights its flexibility, which is an important quality for the de facto language of science, aviation, international trade, etc. So to answer the question, I think there are multiple aspects which make a combined word more likely, including: 1) the constituent words are different parts-of-speech being combined into a new part-of-speech, 2) a contraction whose spelling happens to be very similar to an existing word but is still mostly unambiguous, and 3) how likely the combined word appears in colloquial or subject-matter specific speech and writing.

Even a combo word that meets all three of those points is not guaranteed to be universally accepted as a new word, but some combo words get accepted even if they don’t meet most of those points. Starting with your examples, “greenhouse” certainly meets point #1, since “green” is an adjective and “house” is a noun, which combined form the new noun “greenhouse”. “login” does the same, although it also meets point #2, since it’s a contraction of “log in” (v.), meaning to sign into a web service. The same goes for “work out” (v) and workout (n, an exercise program for each day of week).

I personally always write “log in” when I mean the verb, and “login” (n.) to refer to the credentials needed to sign in. But that’s my CS degree showing. As such, I’m of the opinion that “login” as a verb is a typo, since it’s the wrong part-of-speech (a noun when a verb is needed). Same goes for “alot” (IMO, a typo) and “allot” (v, to allocate). Whether such typos are permissible depends on the quality of the writing, as “takecare” and “ofcourse” would be unacceptable in a dissertation but perfectly fine for an IRC chatroom for die-hard fans of British cooking shows. But I think most people in the latter group, if asked, would probably agree that “ofcourse” is a typo. Basic English literacy means we’re not going to hapazardly throwawaythewhitespace andtryto pretendthatitdoesnt matter.

Finally, prevalence, which is considered by many linguists to be the ultimate test of neologisms: if people use it and it’s understood, then it’s a word. But that rule needs to be viewed from the lens of the intended audience. For example, years and years ago, I understood the legal term “housecar” in the California Vehicle Code to mean something akin to a family saloon car, meaning a car suitable for transporting a whole family or household. This would contrast with a pick-up trucks, sports cars, and commercial vehicles. But my naivete was to not look up the actual definition, and I just wrongly assumed that definition because it made sense for “house” and “car” to combine in that way. The real definition is a car that is also a house, meaning an RV or motorhome. That just goes to show that – since I’m not a lawyer – I was not the audience to gauge whether “housecar” is a valid combo word or not. Whereas California-based lawyers would have likely recognized the meaning in short order.

Basically, each combo word is unique in its circumstances, but perhaps those three points I mentioned have a lot (haha) to do with whether a combo word achieves universal adoption. “alot” does not seem to have met the full acceptance test.