I've met Marcia Angell (along with George Lundborg and Drummond Rennie) at the first Conference on Peer Review in Biomedical Publication. Her worry was that the pharmaceutical industry exhibited too much control over drug research. As time went on, mechanisms to address this have been created, such as trial registries (such as clinicaltrials.gov) and requirements for data submission and code submission. COI has also begin to play a big role. And Cochrane has sued European industries for failure to publish negative drug results.
Listen, if you ever have a paper retracted, your career is over. Period. If you are accused of fraud, no university will want you, and no grant agency will give you one. In general, you do not get a second chance here.
You ask 2 provocative questions, including one that has many questions embedded. Some of the issues you raise are not fraud, but they are certainly scientific misconduct, and depending on what your ICD says could put you and your university or organization at significant risk (i.e. putting active product in placebo mix- since the ICD would need to tell you about the potential risks and the manner in which you are randomized). Ghost-writing is a difficult topic- authorship in scientific publication is well defined- see the definition from the International Committee of Medical Journal; Editors, but with a contract, you are paying someone to write for you and they give you their rights- so while the ethics are questionable- and I don't like this- there is nothing illegal there.
As to your second point, you cannot prove the unprovable. We have never found bigfoot, so he or she or it has never been discovered, but people are still out there looking. The fact that some fraud is not discovered- and this has to be true, of course- is not any incentive to try on your own to commit it. The risk is far too high and honestly, it is never seen with younger novice researchers.
All these issues are not new in scientific publication. We editors grapple with this entire constellation of challenges all the time- and now we have ChatGPT,new forms of literature reviews (living, scoping, narrative, systematic, meta-analysis, and more), publication bias in situ, selective reporting of data (data mining), predatory journals, etc.