In part, this intermittent series of posts will deal with reasoning and knowledge in linguistics, when applied in such an unusual way as conlanging is.
One notion that forms part of the backbone of conlanging thought is the idea that we can just apply reasoning at a very basic level to reach conclusions about typology.
Consider, for instance, pro-drop. Common wisdom is that pro-drop and verbal marking for subjects (and possibly objects) go together. Superficially, this seems reasonable, but we know there are languages that have subject congruence, but do not permit pro-drop. Likewise, we know there are languages that have pro-drop, but don't mark their verbs.
Common wisdom is that lack of case (and/or verbal marking) forces word order to be fixed. But many languages without case marking permit some amount of word order rearrangement - Swedish, for instance, permits both SVO and OVS, without any explicit marking. This to the extent that I have been in situations where people have parsed what I have said (SVO) as OVS, because they have parsed contextual cues and salient features of the words involved in the utterances differently than I would have expected.
Yes, of course Swedish doesn't have strictly free word order - SVO still probably accounts for at least nearly the majority of utterances, followed by AVSO (where A = adverbial), followed by some oddities like VO (with omitted subject), followed by OVS fairly far down the line. The point is, you don't need the case marking to free the word order, what you need it for is to obtain very free word order, that is, word order where the different orders don't significantly differ statistically, and thus make it hard even to learn what is what.
The point I am trying to reach is that ultimately, we cannot rely on ideas like "IFF X is marked in one way, then X can be left unmarked in other ways". Some languages simply structure their utterances in ways where who or what the subject is is irrelevant. In some languages, discourse tends to focus more on events than on people involved, in some languages the discourse is more interested in the who does what aspect of it. Much like some languages don't have tense. Further, with subjects and objects, oftentimes there is a significant bunch of additional knowledge the speaker and listener can be assumed to share, and this makes looking at whether the subject can be retrieved from other markings with regards to pro-drop, or whether the subject can be resolved from other markings with regards to case.
Thus, when we reason about language, we need to acknowledge that the actual form is not IFF X, then Y, but rather if any X out of a huge bunch of unknowns, then maybe Y.