Delivery Channels That Actually Stick

Your automations kept their delivery settings. Now they actually do.

Here's what was happening: you'd open an automation set to post in Slack — or as a GitHub PR comment, a Jira comment, a Linear comment — change one unrelated thing, hit Save, get a cheerful success. Reopen it and the delivery channel had quietly reverted to the portal feed. No error, no warning, nothing to grep for. Just an integration you configured, gone, and reviews landing somewhere your team wasn't looking.

The cause was buried in how partial edits were saved. When only some fields change, we write only those fields — sensible, until the save path went one level too deep and dropped the tag identifying which kind of delivery channel you'd picked. A channel with no type reads back as the default, and the default is the portal. So Slack became Portal on the way to the database, and the API had every reason to think it succeeded.

We replaced that save path with a single shared helper: it still writes only the fields you touched, but it writes their full contents — nested settings, type tags and all. Same treatment now covers automation runs, user profiles, and organization settings, where the same trap was sitting quietly unsprung. Run history keeps its per-action detail, and org settings hold onto deliberate empty values like an unset schedule day instead of silently reinterpreting them.

Edit an automation as many times as you like. The channel you chose is the channel that stays.