In theory. It's a wacky system when you think about it and there's plenty of places it can fall down. But as built, I don't think it's the prosecutor's job to decide innocence or guilt.
They do need to follow the rules, though. If they're obtaining convictions by violating the rules, they need to be punished strongly. If they're obtaining false convictions within the rules then the rules need to be reexamined.
A prosecutor has the full weight of a government and the associated resources behind him. To bring those resources to bear on someone they know is innocent should be a crime all in its own right.
Instead, I'm fine if every stage and every actor in the system was biased towards a finding of not guilty.
To that end, I do think it's incumbent on the prosecutors to only seek to convict those it believes are guilty rather than those it believes it can convict.
Because this helps society?
A prosecutor trying to optimize for conviction rates will still refrain from prosecuting people when he doesn't have a good case. It's just based on the quality of the case, not the presumed guilt or innocence of the accused.
My point of eyewitness testimony is that it would truly convince the DA but wouldn't be valuable evidence in the trial. They could say "I think I saw him elsewhere" to be truthful but say it in a way that didn't give it enough weight to counteract the other, conflicting (because you weren't really there) evidence.
If this came out, the DA would not even face censure or reprimand, let alone firing or criminal charges.
I don't know the hand-ranking of various legal maneuvers but I am curious where a DA's eye-witness testimony would rank compared to a cell-phone location trace, or hair samples, etc.
There isn't a way I can think of to correct it ex post facto, but we can definitely prevent these kinds of problems better. One big area where we can improve things is on data collected.
Currently, the only data I can usually find on prosecutors is usually their conviction rates. But the job of a prosecutor isn't just to convict people, it's to convict people who are guilty. Highly visible conviction rates mean that a prosecutor is incentivized to try only easy-to-win cases, even if the defendant in an easy-to-win case is innocent or the defendant in a hard-to-win case is guilty. We need more effort put into overturning convictions generally, and careful collection data on overturned convictions and that data needs to be publicized equally to conviction rates. If 5% of a prosecutor's convictions are overturned either through appeal or reversal, that should be a huge red flag.
The other thing is, the vast majority of convictions are of poor people who are defended by a public defender. Public defenders are underfunded, understaffed, and as a result, often underqualified (if you're a great lawyer, there's no reason to become a public defender when you can get paid far more for easier work doing prosecution or private practice). Who knows how much of this false evidence would have been discovered earlier if a competent, well-funded, and well-staffed defense lawyer had investigated the evidence? At the very least, I think prosecution budgets should be tied to defense budgets: there's no reason prosecution should ever be funded more than defense. And it might be worth investigating putting public defenders through the same election process as district attorneys (I'm not sure this would work, but I'm not sure it has been tried).