Why your worst case is worse than it looks
What does your open book lose if everything goes wrong at once?
Add up what every open market loses if it resolves the worst way it can. That total is your floor, and it is correct at any correlation.
What correlation changes is how often you get near it. A book of independent bets almost never resolves all-worst at once. A book riding one game script does it routinely — and at ρ = 0.45, twelve positions are 2.0 independent bets, which means "all of them going wrong together" is roughly as likely as two bets going wrong together.
The bettor who sized for twelve independent positions and is holding 2.0 is not slightly over-exposed. They are holding several times the position they think they are.
Two ways the floor itself is understated
Concentration. When more than 35% of open money sits on a single event, the book is that event wearing a portfolio's clothes. Worth a flag on its own, separately from the correlation.
Unknown outcome sets. This is the subtle one. If nothing recorded what results a market can produce, the only outcomes visible are the ones you bet on — so every result the model can see is one you backed, and the "worst case" comes out positive. A one-sided book looks risk-free to any tool that infers the outcome set from the bets in it. The honest response is to say the floor is optimistic and why, not to print it.