A scheduled post went out twice on Facebook and Bluesky. Its first send was cut by PostProxy's gateway on every network, the row read failed, the team made the post again, and the first two networks had taken the first copy after all. Now every send goes through one check: an account that another post of the last 72 hours already carries the same text for is refused, named, and the other accounts go out (a second copy on purpose needs an explicit allow_duplicate). And a send whose every network was cut no longer reads failed: it reads processing with a sentence that says PostProxy is being checked for up to six minutes and the post must not be created again. Also: a lookup on PostProxy never adopts a post another of yours already holds, the sync keeps going when the database connection drops after a long wait, and Pinterest first comments stop being attempted (PostProxy refuses them).