Let's for a moment consider the deponent verb. This is a verb which lacks morphologically active forms, despite being active. This might seem a bit weird, but let us have a look at some Swedish deponent verbs.
First, Swedish has a morphological passive, mostly formed by affixing -s to verb forms. (Swedish also has two periphrastic passives, but this is irrelevant for now.)
Here are some verbs which never appear without their -s:
andas (to breathe)
hoppas (to hope)
minnas (to remember)
låtsas (to pretend)
brås (to take after, to be similar to someone - in both cases due to family connections)
Some of these can take objects (granted, a minority). Andas can take the gas which is breathed ('breathe air', or, say, the aliens of Jupiter breathe methane - varelserna från Jupiter andas metan). "Hoppas" can take det ('that, it') as its object, signifying 'I hope so' (but literally 'I hope that'). Minnas can take any person or thing or fact as its object. Låtsas often is an auxiliary with a transitive verb under it.
In Swedish, these lack a past participle - but some do have a gerund (that morphologically looks exactly like a past participle; however, syntactical differences clarify that it indeed only is a gerund). I will warn against looking into lists of Swedish deponents, because some of them do seem to be just passives with slightly odd semantic shifts, or sometimes even just ... passives. The Swedish -s form also imho is not just a passive marker but also happens to be a reciprocal and an aspectual marker.
Other languages with deponents may have other restrictions - maybe all the deponents are intransitive, or maybe a verb is only partially deponent (i.e. deponent in, say, the participles but not in the finite forms).
Let's make up a set of features:
+ active syntactically
+/- transitive
- active finite forms
- active infinite forms
+ passive finite forms
+ passive infinite forms
- can take agent adverb (e.g. the 'was seen by us' part)
Let's use these features to consider the antideponent.
Verbs such as 'boil' in English seem to permit somewhat similar behaviors, i.e. they can be passive in meaning (or active), thus passive syntactically is partially true. However, English does have active finite and infinite forms for boil, i.e. 'to be boiled' and the participle 'boiled' itself. The active form, 'boiling' interestingly enough does serve to convey the passive meaning of 'being boiled' as well. It cannot, however, take the agent:
the egg is boiling by me is wrong, I'm boiling the egg is acceptable.
Let's inverse the above table fully:
- active syntactically
+/- transitive
+ active finite forms
+ active infinite forms
- passive finite forms
- passive infinite forms
+ can take agent adverb
The interesting bit here is the +/-transitive, and I think that's where we could distinguish this from run off the mill split-ergativity, where some verbs just happen to have an ergative-like behavior. If we restricted this so it only ever happened with intransitives, and the actual subject was demoted to agent adverbial, whereas the subject either is empty or a dummy pronoun, this is getting us into some interesting ground.
Another option is just simply having these as a sort of lexical restriction: these verbs just don't do passive. I think English maybe actually might have some of those even beyond the auxiliaries?
A further option is of course to take something like English 'I broke the window' but only permit these two options:
The window broke.
The window broke by me.
*I broke the window.
Once more voices are involved, some interesting options emerge, such as gaps in the voice paradigm for verbs.
