What's the name of this literary device?
The sentence is a synecdoche:
A whole is represented by naming one of its parts (genus named for species), or vice versa (species named for genus).
(From Silva Rhetoricae, synecdoche, emphasis mine.)
The fragment is a merismus:
The dividing of a whole into its parts.
(From Silva Rhetoricae, merismus.)
However, guessing from my rather glancing acquaintance with you and the terms you're using, I suspect you (or we) are down the rabbithole, which is to say, to borrow a metaphor from my quoted source, Silva Rhetoricae, that you (or we) are becoming lost in the forest of rhetoric. If necessary, I'll justify the claim by noting that paraprosdokian is not to my knowledge a term from classical Latin or Greek rhetoric, but is rather
...a clumsy, malformed, awkward semiliterate neologism.
(From Bill Casselman's Words of the World.)
No offense. To be fair, totum pro parte ('whole for a part', where the silence of the whole theater represents the silence of the spectators), seems probably to be an equitable division of synecdoche, the other half being the pars pro toto ('part for the whole'), even though the gussying up of both with Latin phrases seems gratuitous.
Having been fair, I'll only lift a russet eyebrow at pataphor before hying off with Fergus "to pierce the deep wood's woven shade".
Paraprosdokian is the perfect word to describe what you've pointed to.
The phenomenon you've isolated is definitely a species of paraprosdokian, understood as "a figure of speech in which the latter part of a sentence, phrase, or larger discourse is surprising or unexpected in a way that causes the reader or listener to reframe or reinterpret the first part" (here).
But I also agree with you that you've isolated a finer-grained phenomenon, a paraprosdokian that subverts an initial figure of speech (what you call the totem pro parte). I highly doubt there is a specific word for this, much less a need for one.
Subverted metonym is good for your particular example, which involves a metonym, but it doesn't name the phenomenon of subverting a figure generally. Maybe you could call the general device a figure-subverting paraprosdokian.