Good detectives know the limits of the evidence. But one key is amount of evidence present. If the murder weapon is found in your possession, your finger prints are present, your DNA is present, you were seen in the area at time of the crime, etc. you now have some explaining to do. One item is likely not enough to get a conviction.
Also, DNA testing has improved tremendously since it was first introduced to the point it is one the best methods to positively identify a person (except for identical twins) available. Many cold cases have been solved because of DNA evidence.
The point of the scientific evidence (all evidence actually) is not that each bit proves absolutely the guilt of the person but that there is enough evidence that beyond a reasonable doubt the person is guilty. If you apply probability, if each bit is 95% certain the more bits will push the overall certainty much higher.
The prosecutor's goal is often a conviction but in some jurisdictions the jury is charged with finding the truth.
Back to the issue in the post, part of the problem is crime scene evidence is often poor quality (smudged fingerprints, partial prints, etc.) that make accurate identification very difficult. Couple this with very sloppy evidence handling procedures (OJ trial when the detective kept the samples in his coat pocket) and there is no valid evidence.
The whale oil industry was in trouble when the petroleum industry started. 8-track tapes were made by the same companies that made vinyl records, cassette tapes, and CDs and most of them are still around.
The donut eaters would rather sit around the table eating donuts, drinking coffee, and watching whatever the shop is streaming on the flatscreen. Actually get off their behinds and use the information they already have is a sacrilege called working.
The metadata such numbers called or texted should be enough to allow a brief visit with an abbreviated version of 20 questions. Even if it was a burner phone, a little bit of actual work might tie the phone to a person; it had to be purchased by someone, somewhere.
The whole attack on Apple and encryption smells of a cover up of a FBI failure to break up a what was probably a known terror plot by them. The perjury is in lying to a tame judge that they needed to decrypt his company issued phone.
The emperor lives naked in a glass house. Better learn to be discrete.
I have often wondered if anyone has bothered to read the Constitutional authorization: "To promote the Progress of Science and useful Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries." It seems the key words are "To promote the Progress of Science and useful Arts" and "limited Times". It could be argued that the current patent and even more the current copyright regimes are unconstitutional.
OTP is considered the most secure method with a couple of caveats. There is a maximum length for each message. Each pad must generated in a way that does provide clues into it or others. The pads can never be reused.
Or the correct nickname for the NY Times: "The Grey Whore".
With burner phones, encryption is less critical because it is difficult to trace the phone to specific person. By regularly buying and ditching burner phones and careful sharing of the numbers it is very difficult to tie all the less ends together.
This smells of the DOJ surrendering. They lost key elements of public opinion and pretty much anyone who has fair understanding of cryptography. They also made it imperative the tech industry should push back in Congress which could really make their lives deservedly miserable.
There are several problems with academic publishing. First the publishers often charge the authors a fee to publish. Second the authors are required to surrender their copyright to the publisher. Third the subscriptions and charges per article are unaffordable for most individuals. Fourth, often the research is sponsored by a government agency thus the taxpayers. All are troubling but the last is most.
Also, the idea of publication and peer review is to get results and theories out to the wider academic community not to make a publishing house rich.
As far as breaking codes, natural languages have sound thus letter patterns in the language. If the encrypted text is long enough, the text will have essentially the same distribution as the language. Also, many encrypted messages, particularly government/military tend to have an internal structure much like a business letter. Both of these patterns give clues allowing code breaking. This was observed in the 1920's by William Friedman.
Also, the strength of cypher system is based on the underlying technology used to create it. Modern computer based systems are mathematically very strong for our current technology but they can be broken if one has enough computational power and enough time.
Back to San Bernardino, there is another rule about intelligence: it goes stale the older the information. Even if there is intelligence information on the phone, as it ages it will become much less useful overall.