I agree with WanderingWinder that luck is often an important part of a game's design, and adds more than it takes away.
I disagree that this pertains to 5/2 starts though. Player 1 does not make any decisions before he finds out whether he will open 2/5, 3/4, 4/3, or 5/2. That's a very key difference.
See, Treasure Map adds to the game because before you buy the Treasure Map, you make a decision. Do you buy Treasure Map and take the risk? Do you skip it and not take the risk? Then this complexity enhances the game. The player has to weigh the odds of Treasure Map's failure against its success before he decides whether to buy it. After he buys it, it will either collide or whiff, and will either validate the player's intuitions or punish his foolishness.
When player 1 draws 5/2, he makes no decision before hand whatsoever. His first decision of the game is how many of his five coppers he will play, by that point, the slot machine has already spun. There's no decision that we enriched with the possibility of 3/4 or 5/2 that occurred beforehand. No decision occurred before hand.
Of course, the possibility and mystery that player 2 drew 5/2 or 3/4 actually adds a little depth the player 1's turn one buy. Not enough that I would support player 2's ability to draw either one, but there is definitely a position you can take there that that is good for strategic richness.
But the question of whether player 1 draws 5/2 or 3/4 doesn't enrich any decision. It increases the total number of possibilities in the game, which decreases the likelihood that any two games play out in a similar fashion. It doesn't improve the individual quality of any of those games though.
All that's left is the notion that no two games feel alike. But that's dominion's strength. It's brutish knockout strength. Games feel so dissimilar in Dominion that the community doesn't even put very much time or effort into developing new kingdom randomization rules, because you actually can throw them in a fishbowl and yank cards out and it's hard to improve upon that, every game already feels different. Well. The ones without Governor. And 5/2 mirrors and 5/2 versus 3/4 games have a tendency to feel a heck of a lot like eachother, so much so it's not even obviously clear they are increasing that feel of dissimilarity. 5/2 starts put a lot of fast BM strategies like Witch, Mountebank, Trading Post, and Vault over the top, decreasing their need for other cards because those 5$ cards are designed for the other 86% of games or whatever the percentage is.
I find it incredibly hard to argue that the game is at its best when player 1 might draw 5/2 or 3/4. I find it exponentially easier to oppose any other watering down of randomization, even up to leaving player 2's split random.