Thanks for watching and for the very helpful tips!
My problem is that I generally pick the music after I shoot the video (or at least most of it), so I am cutting existing footage to the music, not the other way around. That means I often have to change the speed of the footage to make it fit the tempo of the music. I agree 100% that those course corrections and speed adjustments are distracting (on the course corrections, I do use a ton of post-stabilization processing, but as you can see, that only goes so far).
For example, that long tracking shot with the full moon on Ventura Beach -- I wanted to cover as much distance as I could and end the shot on the rise up at the end. Could only do that by speeding it up. Same with the shots of the Carpinteria cliffs: I had to make the shots match the drama of the music and the only way (that I know of) to do that was to pump up the speed of the shot. I guess I could have flown the drone faster when I shot it, but again, I didn't know what music I was going to use when I shot it, so there was no way for me to have known that. I guess I need to plan these things out better!
By the way, there is no hyperlapse in this video -- that's just running the footage at 3x or 4x the speed it was shot at. Again, I do this to make the cuts organic to the soundtrack and create as much movement, energy, and distance in the shot as possible. The downside is (as you point out) it amplifies small movements or course corrections.