Oh, never mind - yeah, by using the access logs of segments you can effectively anticipate and pre-encode when you need to. And once the hls or cmaf stabilizes you can just encode one resolution. And the player will tell you it wants to move up or down, so you can trigger the encode it wants.
It's interesting your customers want the video immediately; ours don't care about that. But you guys can really build your manifest files and encode immediately, since you're making those mezzanine files.
Then the encoded files are basically a cache that you can evict whenever.
How long did it take you guys to prove that design out?