Well, how can we get the benefit of wars without the obvious downsides?

Fast feedback loops, obvious win / lose results (a battle takes a few hours, you are dead or not).

Randomly fire members of upper management (equivalent of organisational change by enemy sniper action)?

The research end wasn't really the section getting hit by sniper fire, so that bit probably won't fly.