Don't want to rain on anyone's parade, but I doon't see admirals easily sacrificing a quarter billion dollars just ubecause the unit has been way more successful than expected so far; rather, people adapt their expectations and push for even more. People who roll like this usually don't make it to admiral, and those who do, well, they usually don't gamble. For the final version I suggest a section to explain why Solanta is an exception ... or maybe the girls decide to go rogue, or have exceptionally good arguments.
(But the normal/expected outcome would NOT be that the better argument wins. That's almost never works outside of mathematicians' circles, and even there it would be a hard sell if it weren't for actual proofs.)

Them going rogue is a hard sell too, of course. But maybe HOLMES is monitoring the conversation going bad, and filling them in with background info about the big picture. An AI, unburdened by considerations of social prestige, career, or atavistic hierarchy might be less fazed about a breach of knowledge compartmentalization when a decidedly non-zero chance of global thermonuclear escalation is the stake. The downside to this vector is of course the danger of turning the AI into some stupid deus ex machina (sic) to resolve the plot running into a dead end.