COW forming works great in many cases, specifically if you're drastically shortening neck length - 6 BR to 6 Dasher is an example, +0.080" increase in headspace. My opinion is that it's not necessary in this one particular design because if you spend an extra half hour running a mandrel and setting up your sizing die correctly from the start you get better fit cases after one actual firing than you do from the COW forming shot.
For 243 AI specifically, I use a .2550 mandrel (LE Wilson .257 'new' mandrel) to neck up the case, then fit the case into the rifle using a 6mm bushing neck die to where the bolt was putting more crush on the case than the base AI design does. I ran full-house 243 Win loads for firing forming, and fired one case 15 times without case head sep, losing the primer pocket, or trimming. Stacked five shots in a row on a 1-moa gong at 1200 yards with the un-tuned fire forming load. I used Magnum because I had a lot of it and it's sufficiently slow for my rifle. H4831/H4831SC works well in general in fast twist barrels with heavier bullets. H4350 is probably on the fast side for bullet weights over 100gn because you'll have a lower case fill ratio and for fireforming I want the highest fill rate I can get, but if this is a slow twist barrel for varmint bullets it would work fine.
The advantage of the false shoulder method is the case seals better in the chamber and there was a measurable improvement in
consistency of the formed shoulder over jammed/un-shouldered forming loads. No carbon shooting on the case, no carbon at all past the false shoulder, and calipers don't lie. 3150FPS out of the gate with 105 Bergers
But I shoot loooong barrels, not pencil stubs.
Sheridan Engineering slotted 243WIN gauge to show the neck crush fit in a 243 chamber. In theory a correctly set-up AI chamber would pivot around the neck/shoulder junction, but the false shoulder removes any tolerance stacking and gives you a consistent crush on every case regardless of the virgin brass measurements.