Playback of short Mavic 3 clips pause and are laggy

The file size on disc has nothing to do with playback, really. Or very, very little.

It's the file type ... especially the compression and encoding methods ... that count.

RED media is actually pretty "light" to run, until you hit say 8k, where there's just so flippin' many pixels per frame that the computational load adds up.  Cineform is another codec noted for ease of playback, as are most DNx and ProRes variants. And all of them take some to a lot more disc space.

Drones push the boundaries of both compression and encoding .... like, did you know, there isn't actually a real frame for the vast majority of any of those clips? There's only a complete real frame, if highly compressed, every maybe 15-30 frames of the clip. In long-GOP land, those are called iframes.

Between iframes, you have p or b "frames" ... they are only a dataset of pixels that will change before the next iframe, or that have changed since the last iframe, or ... both.

So the computer has to de-encdode and de-compress up to 60 frames,  and store those images somewhere,  simply to show the next frame on the monitor.

If all an app does is play video files back, the computer can keep up normally. BUT ... in an NLE, like Premiere, Resolve, or Avid, running the app itself is a load, besides playing back a jumbled sequence of bits of clips on a sequence with whatever other effects have been applied.

So unless your computer has the (mostly Intel) bits for CPU/mobo work on long-GOP decoding/decompression, H.264/5 playback is a right bunch of goo.

I work for/with/teach pro colorists. Most of whom 1) run "heavy iron" computers that make our rigs look pathetically weak and cheap and 2) work in Resolve or Baselight. So yea, I'm in Resolve a lot. And in a ton of workflow discussions with the people working from music vids to major movies, docs, and a ton of episodic.

Most colorists immediately transcode any significant mp4/long-GOP media on receiving the client media. And grade using the t-codes. Because they simply don't wanna mess with long-GOP.

And note, this does inlude even some AVCHD clips. 

And there are shops out there, that like has been the process in Avid, that on receiving client media ... t-code all of it to the same codec. For ease of playback and complete similarity between files.

There are still many long-form projects where one of the DIT's requirements ... or at times the assistant editor, depending on how the project is organized ... this person takes all camera media, and transcodes to the specified media for the project so that all media is one format. Then them make proxies from those t-codes, so that any editor can choose at any time whether to run the full media or t-codes.

H.264/5 will be used to send things through frame.io or other services to people who need to see current status of a scene or sequence. But that's about it.

For playback issues with H.264 media, especially from drones, making Cineform or ProRes proxies is typically a wise solution.

Neil