Two separate clips will require extra operations during playback. A simple split or copy will also require extra operations but not as many. While these do not seem like a big deal, they add up and as load increases it means you are that much closer to dropouts with all those extra operations, especially when you are trying to keep latency low enough to record. I am not quite good enough to go much into these implementation details, I just recently got to the point of realizing how important these seemingly minor details matter when it comes to real time audio.

I don't know much about video but, video editing is simpler, you probably have a GPU to take a good chunk of the load off of the CPU and latency is less of an issue. How often do you do you have two dozen video tracks playing at once with each going through a dozen effects while trying to record a few more video tracks in real time all on your CPU?

I tried the test in Logic Pro, and Logic can split, join, reverse, all without creating any new files. It seems this varies per DAW.

> I don't know much about video but, video editing is simpler, you probably have a GPU to take a good chunk of the load off of the CPU and latency is less of an issue. How often do you do you have two dozen video tracks playing at once with each going through a dozen effects while trying to record a few more video tracks in real time all on your CPU?

I wouldn't characterize video in general this way (although NLEs maybe), e.g., a couple of dozen layers with effects in Adobe After Effects is common, and After Effects utilizes more CPU than GPU generally. After Effects can do all these operations without creating new files as well (although it has extremely aggressive caching both in RAM and on disk). (And real-time isn't as much a priority for video of course.)

I think this is partially an implementation detail, I think Ableton probably creates the extra files to help a bit with live performance since that's their focus. Whereas Logic Pro is a more generalist DAW. There are options to deliberately create new files, any time you want, in both (same with NLEs and After Effects).

Right, definitely implementation detail. In a DAW meant for the recording studio there is not much editing, you are dealing with full tracks for the most part and massive amounts of storage for all those takes; creating files is not much use here but in such a DAW you will hit the wall faster if you do a great deal of editing. In a DAW meant for production work it is fairly common to do a lot of editing and assembling tracks from many bits and pieces, creating files becomes more useful and gives you more headroom.

I was talking about NLEs, not video in general, just meant that I am not well versed in video in general, just dabble on occasion. After Effects and layering is quite different, probably can get away with 100ms of latency there without issue which gives a good deal of headroom compared to the ~5ms or less we need for a DAW.