Only listen to targets update

This reduces the complexity of the reducer, as well as how frequently it's run. The VM will emit a targets update after a step if any target info has changed.