I'd think that the main use with it is timing.

A lot of people get into these autopilot patterns where they'll stab, dash a short circle back to the same point, then stab again. For more experienced autopilot players, they're timed to dodge and counter an immediate counterattack, but if you delay your counterattack by midstepping/holding a light charge, you can intercept them while they're still dashing and block their attack input.

What you call 'midstepping' is charging (from a non-final attack) and then chargewalking, then releasing to do a normal attack instead of a charge attack, right?

By that definition, you're delaying your next attack (or as you said, if you wait too long, starting a new combo) rather than just directly starting it, are you not?

edit:
And in fact, you mention it yourself:
Midstepping does however take more time than not doing it

My point from the previous point was just that after you hit someone, they have a tendency to assume you will combo, then backstep and immediately come back to counter that combo hit. The delay that takes "more time than not doing it" is the time that you can use to space and do a counter-counter-hit.

imo you guys were originally agreeing on essentially the same tactic prior.

The essential question, far as I can tell, is whether or not delaying your second attack (be it via charging or just walking) is a viable, productive tactic in t3 and where its use is preferred by the community v just autopilot 1337killzobot. If my opponent is using the same weight of weapon, it's viable for combat in which I can predict the overall timing of the opponent due to them still autobotting. If they're using a lighter weapon, I prefer to fall back to lure them into me as while they would have initiative, their shorter ranged weapon would fail them.

Essentially that's the whole point of this tactic in a nutshell, it says directly to your opponent "I can predict your carp, change the pace please or suffer boredom". That's the time I'd use this tactic, is when I can reliably predict that my opponent is going to autopilot.

I do. It is really fun.

I make decisions about "midstepping" in the same way I make decisions about fast I perform each rush. By default, I don't midstep but at certain times it just happens as a reflex to the behavior of the other people (there are also times when it happens by accident).
Usually I don't midstep when doing ambushes or against opponents I know I can defeat easily (which of course is everybody since I'm OP). In a prolonged battle, I tend to midstep randomly to change my attack pattern. As stated before it's not a conscious decision, more like a reaction to the state of the battle.

Coolest thing to do with this move is adding an extra spin to your attack. Looks sick and is the only reason I would ever use a heavy sword.

I made it a point to watch people who use this in LD.

In the thick of battle when there is too much going on and too much multitasking for you to be noticed (a big group fight) it MIGHT be useful..
If you're an opportunist or a patient player, the first thing you look for is someone who is open to attack or misses their swing so you can punish them.
If I see someone charge a DA or GF mid fight... all you really have to do is either wait till the charge flash (which would make them committed to attempting to use it), or the 2nd swing to hit air since you could easily maintain distance. While they're deciding what they want to do, you could riddle them with bullets or dance around them to lay on a free flourish hit or 3.

But it's one of those things... it has so many bad drawbacks it can play mind games on an opponent. Damage hunters might think you're a victim of lag as you fake/walk into a wall for a second or two while charging. They'd run in for that delicious damage just to eat a standing GF attack to the face.
Feigning weakness is legit if it gets you the point.