"Their value comes from being able to track vehicles and people of interest."
I would suggest a more nuanced characterization. At the moment the information is collected, at least for the stated application of evidentary collection after a crime has occurred, one does not yet know the "people of interest." "Their value comes from being able to track" all "vehicles and people" across a region, so that seconds, minutes, hours, days, or even years later, anyone with access to this information can establish comprehensive movements of anyone within the field of view. By itself, this information is extremely limited in value because it lacks identification information. However, when correlated with other external information embodying more detailed action and identification, this enriched movement-based information becomes a very robust evidentiary knowledge base.
If sky eyes becomes routine, they will support ongoing investigations by providing near real time suspect movement information as it happens. "Tailing someone" as we know it may become obsolete, supplanted by strategically placed collection assets dynamically vectored in to illuminate the detailed activity that occurs at suspect stops.
One simple example available today is the extensive network of law enforcement and commercial ALPR tools, to augment the anonymous but near-complete sky eye movement database. Automated correlation of just the ALPR identification spot data with the sky eye comprehensive vehicle movement data can provide a very impressive capability. In isolation, each tool has significant limitations, but together they can be profoundly more capable. Now extend this thought to include confirmation and correlation with other unique identification systems, such as credit card records (location at a very rough time in a specific location), EZpass, consumer RFID tags in your garments, tires, etc. The possibilities seem endless. Even error prone systems can still add value (think current state of the art facial recognition systems) when multiple sources are correlated and filtered for consistency.
The importance of automated consistency enforcement cannot be over-appreciated. Any system has some uncertainty in time and space which sometimes might appear to limit its value. For example, vehicles passing an ALPR site at speed might be mis-associated with adjacent moving contacts in the imagery track data. However, this uncertainty evaporates over time if multiple hypotheses are carried forward, because after vehicles on separate trajectories pass a number of different fixed and mobile ALPR systems, automated systems can deliver unique identifications with high confidence.
The critical innovation is the change from intermittent to continuous observation. It is a challenge for anyone to piece together the many sources of information from brief interactions with a suspect. Currently, it is inevitable that pieces of available evidence will be missed because they will not be imagined. What law enforcement official would know that some significant interaction might have been captured by surveillance at the local Walmart, if, having only intermittent surveillance information, they do not even know that the suspect stopped there? One can imagine that complete, continuous routing and destination information can be a profound enabler for both ongoing and retrospective investigations.
Beyond that, ever greater computing power, algorithms, and automation can ultimately lead to the case of full dynamic information collection and correlation for a comprehensive information reposititory. Such an integrated database of every person's activities could subsequently be parsed for elements of interest well after the occasion.
On the one hand, since such a comprehensive record will inevitably outstrip one's personal memory of one's own actions, one who is honestly mistaken might easily appear to be dishonest.
On the other hand, you might not care to run for political office in such a world. It seems plausible that one's entire life from birth to the present may be "scrubbed" by your in competitors, in hopes that they might find any trivial piccidillio which they might publicise to lead to a more favorable outcome for themselves, and a less favorable outcome for you.
Naturally we imagine such a resource guarded by robust access controls managed by wise and benificient beings for the greater good. Let us hope it is so.