Most devices that will play video these days are powerful enough to do the decoding themselves and have the bandwidth available.
IME this varies a lot between devices. Google TV dongles for example, even the 4K versions, are built with extremely weak SoCs (as in early 2010s phone weak) and lean hard on hardware acceleration. If you want to play back a format that isn’t hardware accelerated on one of these, you’ll have to rely on media server transcoding.
You can tell Emby and Jellyfin to direct play. Pretty sure Plex has that option, but I've not used it in a few years so could be that changed.
Most audio streams are 'direct stream' when you use plex/jf/emby as backend, but your receiver/soundbar only gets the PCM stream, without any object data (yes, you loose atmos, not that it is a lot of loss when using a soundbar, but if you have atmos speakers in your ceiling you want that data)
in an ideal world, the appletv will simply passthrough the audio upstream, so you receiver can do what it does best.