So the caster fumbles and suffers no direct penalty. The front line fighter is penalized instead.All this talk of probability of critical fumbles is off topic. Yes, a great swordsman who fails 5% of the time wouldn't be a great swordsman.
But you need to look at it a different way. In combat many things are going on (rarely a one on one affair), it isn't just that the swordsman screwed up and failed (luck plays a part in these battles, good luck for the target in this case) it could be narrated that while his strike was on target for that chink in the armor below the armpit, but his friend the dwarf who is fighting along side him and attacked immediately before cause the opponent to grunt in pain bringing his arms down at the last second causing the sword strike to deflect off his armor instead perhaps being accidentally re-directed into his friend the dwarf. The swordsman did nothing wrong here, but in combat stuff happens that was not planned.
In my games I allow/expect my players to narratively decide what happens on a critical (failure or success). They can choose to be as punitive or not as they see fit.
A few games ago we had a cleric who being spooked after being woken up by the roar of a bear coming into camp, did what he always did when startled, up-cast a guiding bolt before getting out of his sleeping bag. He rolled a Natural 1. Based on his positioning and his paladin friends positioning compared to the bear. He narrated, that his hands got tangled up in the sleeping bag as he cast his spell through the sleeping bag destroying it and hitting his paladins shield from the back side. He also rolled nearly max damage so the shield was bent and ruined and nearly ripped from his arm.
All of this is the players choice. In my opinion it has worked well.
Sounds typical ... ranged attacks rarely harm the one that fumbles. In your other example what if there is no one close?
Anyway, if it works for you, great. I don't want to ever deal with it again and wouldn't play a game that uses it.