Bug 33766

In handling channels, there may be a need to reliably confirm certain facts before the user is alerted by an approver, or the channel is dispatched to the handler.
As an example, before launching an incoming call alert, the following actions need to be completed:
* The logger needs to reliably store the incoming call event;
* The resource policy hook needs to reserve resources necessary for the call.

The message flow described in bug #27860 details how one client can "hijack" the channel, but it does not describe how multiple entities can serve as mandatory checkpoints.
The observer property DelayApprovers looks like the necessary hinge to implement this: it's possible for observers to delay return from ObserveChannels until they have confirmed readiness. If this is deemed to be a good enough way (the channel dispatcher is supposed to give sensible timeouts to ObserveChannels?), it should be documented.