I'm of the opinion spying should just be based on chance. Let's say 10 AS vs 10 DS. When they make contact, each AS has separate chance of successfully obtaining a defense report depending on the number of DS and the quality of AS and DS (else upgrades for spies be pointless). You could even have two rounds of encounter. In the first round, the success of obtaining a report is determined for each AS. In the second round, whether the AS gets away is determined using a separate chance calculation. This way you can have an upgrade for improving the chance of obtaining a report and another upgrade for improving the chance of the spy getting away. Any spy that fails the get-away roll will be caught and a third set of chance calculation is used to determine whether the defending player finds out who sent the spy or if the defending player is just told that someone had spied.

With this system the attacking player can send lots of spies in hope of getting a defense report but also increase the chance the defending player will be alerted. Or the attacking player can send a few high level spies to try to get a report without being spotted but since there's less spies there's a chance you won't get a report. In this system, its possible for the AS to survive but not generate a defense report.

I think this system has a better feel to it then just mass producing spies and have them go at each other in their own separate melee.