SCOTUS is a weird world.
I'm reminded of the jury in 12 Angry Men. In it, a single juror manages to turn the jury from a pronouncement of guilt to innocence primarily by questioning shaky eyewitness testimony.
Trouble is that:
a) that was a movie, with a script. Not reality
b) the eyewitness was pretty shaky in that case. (an old woman looking at night for a mere flicker of an event)
c) that was 50 + years ago and despite that, people still believe the emotional power of eyewitness testimonies to be roughly infallible.
Which is, in the parlance of evidence, bullshit. Most of the time eyewitnesses are unreliable. Their stories don't match. Police behaviors and even common police techniques (lineups for instance) can manipulate their memories. Even after previous court rulings have aborted some of the most ridiculous manipulations by police. Most people don't have ironclad memory and details will shift over time (I have only recently discovered this as a problem by observing others. My memory seems to work differently, though not perfectly either). Details occurring in rapid succession for complex events like a robbery, rape or murder are often stashed as peripheral data anyway.
The court's position that other unreliable data points would be up for grabs is perhaps disconcerting for the amount of shift that would be required in our criminal justice system. But there's plenty of tactics and methods that are in fact, roughly useless. Drug dogs for instance give false positives at least 2/3s of the time. This isn't the dog's fault. It's the handler. This results in a lot of invasive searches being "justified" when all the justification amounts to is the officer handling the search simply thought the person was guilty, without any evidence to suggest it. As the court pointed out, there are plenty of "jailhouse snitches" testifying (Omar's appearance in the Wire is a classic). The reliability of such people can already be deemed questionable by many observers, though there are probably sensible reasons to regard it as more reliable than is commonly believed (snitching is looked down upon enough in the criminal community that one has to have a pretty damned good reason to be willing to do it, this cuts both ways of course).
Now I would say that having people tell stories about what happened, from their point of view, is perfectly fine. I'm not sure that preventing witnesses from testifying makes sense. But clearly there needs to be much more rigor in how we allow it as admissible evidence. Do lots of people all ID the same person? Is there corroborating video or audio, or actual hard evidence (blood, DNA, fingerprints, ballistics, etc)? I'm inclined to agree with the Court that it's hard to keep it out. But
we should also be more aware of just how bad, how very bad, such
"evidence" is. It's effectively useless.
Ultimately, I'd say suspicion of guilt is justified on such things as "I saw that man, with these peculiar features, doing X to person/property X", certainly. I don't see how we should turn it into firm statements of conviction.