Real time contexts are event sets. Events are strictly ordered given a simple locale but may overlap given multiple locales so event order is not the only signature. When identifying a situation semantic in an emerging phase, event order is very important but so is proximity of multiple instances of event types. Say you have a system of sensors, you may be tracking the order that these send alerts to determine time and speed of some object. You may be monitoring multiple sensor clusters and correlating these values in a higher dimensional space.
To identify an emergent event and dispatch on warning (roughly analogous to launch on warning), you have to be able to assign force values to the events because the emergence of a situation may be presaged by weak signals in combination with strong signals. Proximity is a scalar for weak-to-strong events.
Proximity is one of the signals that you monitor given a locale ontology to make a threat assessment. In an intelligence assessment, you may have multiple intelligence types and fusing these into actionable information where the responder assets are limited is the key to effective dispatch. Otherwise, all you are doing is running to false alarms although false alarms are also a key piece of information. Eventually, all of these signals coalesce and are assigned a value from a code list.
The tricky bit here is learning to get all of this information from multiple XML streams emanating from different information source types (think HUMINT, OSSINT, SIGINT, etc). If the control is emergent, then it may just be an Xlink type, at least, that is one way to do it. Metadata can be emergent information, in fact, likely is.