Detail #134: A language with very reduced role marking and reconstituting the role marking in a weird way

Imagine a language in which there is no 'preferred' voice. However, there are any number of circumstantial voices (voices that promote an oblique argument to subject). All non-subjects are marked in an identical fashion. An additional tendency is that indefinite nouns go to the right (unless they are subjects).

Thus, say,

'mom cooks beans in the new pot' -> se mom cooks.active e new pot e beans,
'mom cooks the beans in a new way' -> se mom cooks-active e beans e new way
'the beans are cooked by mom in a new way' -> se beans cooks-passive e mom e new way
'the beans are cooked by mom in the new pot' -> se beans cooks-passive e new pot e mom
'the new pot is cooked-in beans in a new way' -> se new pot cooks-some.circumstantial-voice e beans e new ways

etc.

Contextual information and lexical information (about what types of nouns tend to appear with what kind of role) will thus be very important for unerstanding an utterance. Of course, longer utterances may be easier to parse due to more hints at who is doing what. One could imaginably also have adjectives that are basically participles-with-voice that co-refer with the verb, and permit disambiguation when needed:

'mom cooks e beans e new in-cooking pot'
'mom cooks e being-cooked beans e pot'

etc.

Here we get an interesting possibility - pro-verbs! Maybe different main verbs take different markers such as this, also permitting us to conflate several verb phrases.

'dad hunts and cuts up and gathers, mom tends and cooks e done-passive* deer, e had-passive** used-passive**** chickens, e taken-passive*** used-instrumental_circumstantial**** herbs
* referring back to "hunts" or such
**referring back to "tends"
***referring back to "gathers"
****referring back to "cooks"

Of course, different verbs have different "verbal class", and thus different auxiliary-participles will refer to them, helping to correlate nouns with the role they have for different verbs. Nouns whose use is expected to be understood without such an auxiliary participle are left unmarked.

At this point, I've basically ruined the idea this post started out as, and built a typologically quite unlikely thing. However, I like the looks of it.