Her petition is timely, and she has exhausted her state remedies (and doesn't have to use state court postconviction proceedings before going federal, because she raised the issue in her initial trial).
Now note that the CA Supreme Court is not bound by 9th Cir precedent.
So new rules don't apply on habeas unless we meet the two exceptions. So we know we don't have the exceptions here, and the question is just: is this a new rule, and is it clearly established? This is pre-AEDPA, so we don't have to just look at SCOTUS precedent.
So our question is whether the subsequent announcement of the proper penalty standard cures the defect, and then whether that rule is new or clearly established.
Standard for watershed rules (second Teague exception):
Could reasonable minds differ about whether this was an established rule, or can we go with the 9th precedent, once we get into federal court, because the district court would be bound by it? If so, are the state courts actually bound by this precedent, in the sense that the petitioner will just go to federal court next?
Chen is maybe smart to go right to habeas, because if she went to SCOTUS after her state supreme court appeal is denied, the court might just want to adopt the 5th Circuit precedent as the rule. But by going to federal district court, the 9th circuit is binding.
But when it then gets to SCOTUS after the federal habeas, SCOTUS can still say that they don't want to make a new rule on habeas, so we're back to the question of what the old rule is. After all, the fact that there's a split in the circuits indicates that reasonable minds differ, and therefore the split in the circuits must remain a mystery, because they can't announce a new rule if the party can't benefit from it.
Note: if there's a dissenting opinion on a case, that might mean it can't be a new rule applied on habeas, because reasonable minds do differ about it.
So the district court is in a pickle: they're bound by the 9th Cir precedent, but the know that if it gets appealed to SCOTUS, they'll get overruled.
Anyway, under Teague, the only rule that's applicable is Alpha v. Beta.
We've had a decision on the merits, and we've got clearly established supreme court precedent, so was the state court's decision contrary to it, or was it an unreasonable interpretation of it?
Stevens finds that "contrary to" and "unreasonable application" are really just one thing: a mood (p. 968). Stevens basically uses AEDPA to undo Teague, which is pretty agressive: we can correct things that are wrong.
O'Connor hews much closer to the text. See the chart on p. 986. Get the statement of the law correct, first off. But it's the "unreasonable application" portion that Stevens has really elided: that second clause has to mean something. The state court can get some things wrong, but not get them wrong unreasonably (i.e. SCOTUS would say this was wrong on direct review, but not on habeas). So things have to be especially wrong to be squashed on habeas (see p. 985, and the Venn diagram).
So the dissenters like this higher standard, but they disagree on the issue of unreasonability (there's also confusion about whether the state court got the rule 100% right-- there's confusion between Strickland and Lockheart-- but we're not going to worry about this). So was it unreasonable for the VA court to say that there wasn't a reasonable probability of a different outcome? O'Connor: yes. Rhenquist: no. The double-reasonability makes this confusing.
How unreasonable does a state court have to be in order to be so unreasonable that it doesn't count as a final result? Or, conversely, what degree of unreasonability will we tolerate on habeas? How willing are we to make state courts re-do their work/deter or incent them, etc?
So we have the spectrum: Stevens is most willing to undo state court results, O'Connor has a higher standard for undoing but is willing to find unreasonableness, and Rhenquist agrees with that standard but is less prone to seeing things as unreasonable.
And the bottom line is that habeas relief is for when you're both wrong and unreasonable. Some people think that's even more insulting to the states, because when they are overturned, they're being told that they are both wrong and unreasonable.
Instead (O'Connor, and also the Rhenquist dissenters, which is enough to be a majority), we follow the chart on p. 986. This is sort of thinking about when we should leave state courts alone versus when we should make them do things over. If they articulate the law correctly, and apply it reasonably, we'll leave them alone.
But maybe 2254(d)(1) is like that: it tells the courts how to interpret a right, and maybe we have an Article III problem. So Congress could completely withdraw jurisdiction over habeas for state court decisions, but is it an abuse of the courts to tell them how to interpret habeas rights this way (i.e., by telling the court to deny relief even if the state court was wrong, as long as it wasn't unreasonably wrong)? This argument has been made a couple of times, but has been rejected at the Circuit level here: this isn't an interpretation of rights, it's a standard of when you're entitled to a remedy.
Williams was an hour away from being executed when SCOTUS took the case-- why would they take it? Not to rescue this one individual, but because the case is a good vehicle for articulating an answer to a question that will be useful in many cases to come.
So, basically, under what circumstances can you develop the facts further on habeas? Generally, the federal habeas court wants to rely on the facts decided by the state court-- it's not practical, generally, on habeas to develop the record further: people are in custody, years have gone by, etc.
We had the standard from Keeney v. Tamayo-Reyes (p. 988), for when deference should be given to the state court. But that was judge-created doctrine, and now we have AEDPA (see p. 1059, middle of the page). This standard is more restrictive: if the claimant has failed to develop the facts in state court, the court SHALL NOT grant a hearing unless A AND B. And for A, you need either the Teague "new rule" exception, or we're talking about facts that could not have been discovered previously through due diligence. And on top of that (B), the facts underlying the claim would establish by clear and convincing evidence that but for the constitutional error, no reasonable factfinder would have found the applicant guilty.
So that's actually kind of like the cause-and-prejudice standard.
And remember that for (B), we may not be talking about actual innocence here-- it might be just innocent of the death penalty. So here, all reasonable factfinders would have found him guilty of the murder. Can we really say that but for the due process error (of empaneling the biased juror), all reasonable factfinders would say Williams shouldn't get the death penalty? Seems far-fetched.
But there's also what "fail" means: if the applicant has failed to develop the facts. Is this like "I fail to understand what you're talking about" or "I failed to fulfill my responsibility?" The way the court saw it, it wasn't a lack of diligence on Williams's part that caused him not to bring these facts out at trial. That's the thing that could save Williams from the A and B problem. The prosecution wants to say that the lack of development is enough to make it so that you have to satisfy A and B (i.e., the "weak" type of fail). If not, though, he'll be able to get out of AEDPA and into Keeny (where he'll still have to show cause and prejudice, but the prosecutorial misconduct he alleges will suffice for cause).
We don't want lots of evidentiary hearings, so we make it tough to get them. But the court saw this as an important enough issue to make it worth crafting some AEDPA doctrine. What was the issue that makes the court unanimous here? What is it they're trying to preserve, given that they know this was a statute intended to cut back on federal habeas jurisdiction, give more deference to state judges, and expedite the death penalty?
It's the fundamental miscarriage of justice prong from Keeny. If we accept the prosecution's theory of the word "fail," then even if defense lawyers do everything right, and the prosecutors do wrong things, they could still get away with it-- we're basically enabling prosecutorial misconduct.
So the court is interpreting the statute to mean what they think it should mean in the context of proper jurisprudence. We want to pick the standard that puts the right incentive on the prosecution.