I am not from Unity, but here is my answer in case useful.
If you are using an Animation Track (the main thing I use when animating characters), I design my characters to all have the same components. E.g. I wrote my own “facial expression” component. I then created little animation clips to set the strengths of these expressions. Each character gets its own blendshapes etc for the different emotions.
I also use “Humanoid” animation clips - which also adds a level of abstraction between the animation movements and the character. After doing that, replacing a character is a drag and drop into the Animation Track binding section. Timeline assets (stored as files on disk) are not tied to characters. You have to create a “scene” and then drop the character you are using from the scene into the binding slot.
I do simple animation inside Unity using their tools. They are pretty basic. Keyframe properties at a particular time. (They do have new “Live Capture” tools, but I am using some free open source tools so no experience with them yet.)
For example, I have a component to make the eyes of a character look at a target object. I then animated via keyframes the Y position of the target object so they eyes went from looking straight ahead to down. Simple stuff, and it works fine.
I would not personally want to create a full body walk animation this way. Instead I find existing animation clips from the asset store and other places. What interests me is also services like DeepMotion.com that you can upload a video clip to and it generates an animation clip for you. There are also open source tools around. They are getting better! (I think a full mocap suit is still best.)
I find it more useful to create animation clips (I am creating a series, so I reuse them a lot - e.g. different character poses, sitting, running). I am experimenting with webcam based (free software!) mocap, leap motion cameras, Quest 2 headset and VR controllers, etc. All cheap things (this is a hobby for me!) Full mocap suits are tempting, but the price tag still too high for me.
The end result is an animation clip or pose that I blend in the timeline. I do some live recordings, but I minimize it so I can produce episodes faster. My favorite tool is using a Quest 2 VR headset that generates Blender BVH files. I am not doing hollywood grade productions, just a hobby, so I can get away with this!