Yes I see now, two.
Well in a perfectly neutral system. Bremmsstrahlung is proportional to electron density, electron temperature, and the ratio of electrons to ion Z. These virtual cathode systems are by definition not perfectly neutral, as the electron/ion ratio > 1 sets up the electrostatic well.
Only to a point, as Nebel suggested with the 'optimum' qualifier, as the power gain function is not linear in all its parameters.
As I understand it, though Rider/Nevins correctly point out the 2nd law issues in play, there are two areas where they fall short: 1) the electron confinement times for a virtual cathode device are shorter than the thermalization/collision time with ions so that the electron temperature never has the opportunity to rise enough to cause unsustainable Bremmstrahlung, 2)their mathematical treatment of collisionality is inadequate. That is, the FP model performed by Chacon et al 2000 improves power gain (Q) by 5 to 10x over that predicted by Nevins. Take this last part up with Chacon et al.