Kadeton,
I agree with much of what you say, especially about it coming down to a clarification. Clarification would show the intent of the rule, and that is something I can not presume to know.
However, this part of your post (see below) I think I've found an answer to:
Kadeton wrote "The question is whether allegiance (enemy or friendly) is conveyed to caused actions, when all other effects are not. There is no clear rule to reference for this - it will come down to a clarification. Either way has significant implications if the clarification is expressed in general terms."
Since I only have the small rulebook I'm using that, but I will quote it directly and completly, with exception to my own typos, as much as it pains me to do so using a tablet.
Pg 23, friendly and enemy call out box:
"Friendly and Enemy
Some rules reference friendly or enemy models. From a mechanical standpoint, friendly models are any models that are on the same Crew as the model the rule is affecting. Enemy models are any model that isn't on the same Crew.
If a rule does not specifically state that it only affects friendly or enemy models (for instance legal targets for an attack) then it can and does affect any model.
When a modle's Actions are controlled by the opponent (for instance with the Obey Action) the model does not change which models are considered friendly, its friends are still its friends."
Allegiance does not appear to regard actions at all, only models (although this thread has taught me that it's very hard not to describe an action that a friendly model takes as a friendly action)
As for the actions causing actions argument, it does not say that actions caused by actions are part of the original action. It states that the orriginal action isnt resolved until the new action is resolved. This specifically states that the new action is both a new action, and another action (another being not the same action).
I too will bow out now since all I could add would be to repeat that:
1) Obey causes a model to perform a (1) action, and does not move a model.
2) the action may be declared as a (1) Walk action
3) the model taking the (1) Walk action is friendly to itself.
Overall I have enjoyed this discussion, have learned a lot of rules interactions from this thread and while researching for this thread, and I appreciate the opposing views posted by you and Rgarbonzo.
Cheers,
GG