Fun fact: HTTP RFCs and drafts between ~1994 and 1999 assumed that servers using compression were storing precompressed data:

"The content-coding is a characteristic of the entity identified by the Request-URI. Typically, the entity-body is stored with this encoding and is only decoded before rendering or analogous usage."

This assumption is critical for interpreting a later sentence that otherwise makes no sense, about servers being expected to default to the "identity" non-encoding... if it is "available." Why would non-encoding ever not be an option? Precompression.

This is non-normative flavor text in the RFC, but I still think it's useful for interpreting its guidance for my weird-ass situation.