A policy that my startup partner set up that really helped a lot: “Nobody is allowed to assume that a teammate has negative motivations”. Unless someone explicitly says something like “I screwed that up because I hate y’all” they should always be given the benefit of the doubt, and all public discourse must assume that they are doing their best and have good intent. Once we got people to buy into that mentality it solved a lot of interpersonal conflict issues.
Another thing we had success with was to encourage people collaborating to interact in public channels more and DM less. That serves 2 purposes really: it leaves a public record on decisions that get made by a collaboration team, but also if there are people on that team who can get caustic, they are less likely to do that when they know other people (including higher ups) can go back and read what they said.