Open flotter opened 2 months ago
Discussed with @benhoyt and will propose a couple of PRs soon.
Plan Manager is notifying all managers of a plan change before the state engine has started (overlord loop). This means managers have no opportunity to remove leftover persisted tasks during manager StartUp
before the manager's change handler starts processing the new plan.
Gracefully handle missing Cached lookups, as this almost certainly means a leftover check from the state engine started running after a reboot. In this case we likely do not have any undo
tasks associated with the handlers in question, so we could simply call Abort on them.
This could be fix for this issue identified above, but I would still like to look at improving the Plan Manager so that it does not propagate a plan change before managers can really cope with it (before the state engine is ready).