Languages throughout the word are dying off. Language death is when no one is raised from infancy speaking that language. So, despite cries of Foul! from dedicated (fanatical?) Latinists is a dead language and though spoken in some families no one lives in a community of Latin speakers and even if, like Medieval monks, they did, Latin would still be dead by definition. That does not mean it would not be lively, just not a alive, i.e. dead.
So we have lost deponent verbs as found in Latin, yet the function of those verbs, a kind of Middle Voice paralleling Active and Passive, still survives with the enclitic “se” in Spanish, for example. The subjunctive is moribund in English but its function is taken up by modal verbs.
OTOH, the ergative may be lost eventually as languages that use it tend to be minor languages with few speakers, e.g. Basque, Australian Aboriginal languages, Mayan dialects (interesting that in this case some Mayan languages so closely related they are considered dialects do not have ergative), and a few others. But Hindi-Urdu does have ergativity and speakers number close to a billion. It is interesting that the same issue rears its head as with the Mayan dialects: so many languages of the subcontinent are very closely related to Hindi-Urdu yet lack ergativity.
But say all languages using ergativity were lost. What might be lost with them? To me, the notion that the subject of an intransitive verb shares properties with the object of a transitive verbs carries an intriguing way of looking at the effects of an action, viz. that the subject of an intransitive verb can be affected in the way the object of a transitive verb is, e.g. John broke the window and The window broke contains a window that is affected by the action of the verb. Even John sleeps shows John participating as a patient or sufferer of the action of sleeping.
I think that is interesting and worth preserving, especially since so many philosophers base their thinking on the operations of grammar.