Hererra v. Collins must be considered in the (unfortunately complicated) context of post-conviction review law. There are state and federal post-conviction review procedures for dealing with new evidence. In addition, while Herrera and later cases reserve the question of whether "actual innocence" is by itself a basis for habeas relief, in practice there is a "bootstrap" Constitutional violation (usually ineffective assistance of counsel) that in conjunction with the actual innocence claim will provide a basis for habeas review. And in McQuiggin v. Perkins, the Supreme Court cleared pretty much any procedural hurdle to such claims when "actual innocence" is shown.

Here is an article that discusses the context surrounding Herrera v. Collins: http://digitalcommons.law.scu.edu/cgi/viewcontent.cgi?articl....

Many (and, I hope, all) jurisdictions will reopen a trial when significant new evidence appears, but not just for any new evidence. For example, "I suddenly remember that, the morning of the crime, I woke up five minutes earlier than usual" or "it turns out there are not 20, but 21 HD videos of the crime" may be new, but is very unlikely to be deemed reason to reopen a trial.

Problem is to determine what is both significant and new.

Even somebody else confessing or new evidence showing up that the convicted has an alibi to the crime may not be deemed significant, as somebody must judge whether the new information is true.

For example, blindly trusting confessors to speak the truth would open the door for groups of people to keep each other unconvicted (A gets indicted and, a few months later, convicted, B suddenly shows up claiming he was with A, giving A an alibi, a year later A gets convicted again because that alibi didn't turn out to be watertight, C (who may already be in jail for something else) confesses, C gets convicted, D, who's dying, confesses, etc.)

So, some judge must determine whether new evidence warrants reopening a case. If that judge is (consciously or unconsciously) biased, this easily can lead to injustice. From what I read, that seems to have been the case quite a few times in the US.

Speaking of pretty clear, later evidence is pretty clear grounds for a new trial according to U.S. law, at least up to 3 years after they're found guilty. Beyond that I'd leave it to the lawyers to figure out.

https://www.law.cornell.edu/rules/frcrmp/rule_33

But that challenges the whole foundations of a civilized society, does it not? I suspect if prominent people are being caught on the wrong side of such procedural issues, ways would be found to tweak the law in a hurry.

Convictions get vacated when an alternate perpetrator is identified or confesses.

Still, that alternate confession must be credible. For high-profile crimes, there often seem to be dozens of alternate confessions, and they are usually obviously untrue so they should not be grounds for appeal of the convicted.

The hard part is how to establish that alternate confessions or new evidence are really significant (as they indeed sometimes are).