Today the way Vera works in general is like “fire and forget”.
This is ok for a single command that is sent to a device because the Z-Wave protocol will assure (more or less) the execution of the command.
Above this Vera provides us with the mechanism of scenes. In scenes we can arrange a bunch of Z-Wave commands. And then we lose the assurance that every command will be executed because between the first command of a scene and the last one a lot of unexpected things may occur. Examples are upgrades, restarts etc.
I like to suggest another model than “fire and forget”:
Vera should use a central table where for every device the desired status and the actual status for every point in time is stored. Scenes and events will not trigger commands directly they update solely this table. Vera sends commands only when the desired status and the actual status are different.
This can help to make Vera more reliable.