So, I'm guessing you're just going on your gut here, and while I can appreciate that, the math doesn't actually support your intentions.

So, 2d6 with 3 Fails and 0 Crits gives you a 25% chance of success at 0 cost.
2d8 with 3 Fails and 0 Crits gives you a 39% chance of success, at a cost of 3 SD (18 Souls) + all future SD cost +2 Souls

2d6 with 2 Fails and 0 Crits gives you a 44% chance of success, at a cost of 3 SD (18 Souls)
2d8 with 2 Fails and 0 Crits gives you a 56% chance of success at a cost of 6 SD (42 Souls)

2d6 with 2 Fails and 1 Crit gives you a 59% chance of success at a cost of 5 SD (30 Souls)
2d8 with 2 Fails and 1 Crit gives you a 65% chance of success at a cost of 8 SD (58 Souls)

2d6 with 2 Fails and 2 Crit gives you 71% chance of success at a cost of 8 SD (48 Souls)

There's basically never a point where upgrading the die size is an improvement for the same cost. Even when you start adding dice:

3d8 with 2 Fails and 1 Crit gives you a 57% chance of success (drops by 8%)
3d6 with 2 Fails and 1 Crit gives you a 65% chance of success (drops by 6%)

And then lets just take a look at a maxed out d6:
2d6 with 1 Fail and 3 Crit gives you a 91% chance of success.

Then a maxed out d8:
2d8 with 1 Fail and 4 Crit gives you a 94% chance of success, but costs waaaay more than that maxed out d6 (Cost of upgrading the size + the cost of the extra crit + all the extra souls required per SD, even if you took the route of upgrading the crits before upgrading its size, that's still an extra 154 souls extra (3 (die upgrade) + 8 (4th crit) + 9 (remove 2 fails) for 120 souls, plus the 34 soul tax on the 4th crit and fail removals).

Then, as we add more dice:
3d6 with 1 Fail and 3 Crit gives you a 92% chance of success. Your odds actually go up by 1 percent.
3d8 with 1 Fail and 4 Crit still gives you a 94% chance of success.

4d6: 93% success
4d8: 94% success
4d10: 93% success

Basically, at least based on the math, given the costs, there isn't really any mechanical reason to upgrade your die size, even in the short term. And since the die sizes aren't tied to weapon type or anything, there isn't even a flavor reason to not go the "unoptimized" route.

Definitely not trying to come off as a jerk or anything. But since it doesn't seem like I'm missing anything, the upgrades might be something you want to address before the final version. Limiting max damage on lower size dice could be enough--if you have to roll the d6 more to kill an enemy, then you've got more chances to fail. Maybe have each die upgrade come with a free +1 damage upgrade on top of that?