Thanks for answers in this thread. I didn't really mean to have YOLO-type random access to production. I was hoping there are ways in between to bridge that gap between dev and ops in those systems, similarly how it has been done eg with SRE in more relaxed security applications. I was hoping for some solutions on the spectrum are adopted more, like mentioned cetralized logs stripped of private data or granting temporary audited access. But it seems with legacy systems this is much harder to implements. I believe there is an optimum balance where actually fewer mistakes could be made if both people developing and operating te system had more visibility into each other field. As for willful fraud attempts, well you can't rule out devs would do it, so of course there should be various barriers preventing that and proper change management, but, my sampling bias aside, when I look at some recent scandals in finance, take eg Wirecard as the last one, there is more often higher management involved than devs.