The trick is "repo", or repurchase agreements.
(1) Buy bonds
(2) Use those bonds as collateral for a low-interest loan
(3) Use the loan money to buy bonds
(4) goto 2
See, e.g [1]
[1] https://www.bloomberg.com/news/articles/2020-04-15/how-repo-...
That means that investors' internal risk limits are the binding constraint, not repo haircuts.
It's another way of saying that the financial sector sets its own leverage. Historically, that has not turned out well. It's why Dodd-Frank included a leverage rule for large banks.
While part of risk, expected return is a larger binding constraint in most cases over risk limits. I'm probably not going to lever up 20x for an tiny expected return. On the other hand, I may very well lever up 5-10x on something 50x more risky than treasuries if the 10yr is yielding 0.725%.
There is some credible research which suggests that large financial institutions act as if they are optimizing mean return subject to a VaR constraint [1].
But for the sake of argument: large bank VaR models affect hedge funds because hedge funds get their leverage through their prime brokers which are... large investment banks.
[EDIT] well no that can't be it because it requires even more money coming in for those loans, which can't provide more expected return than the bonds they're buying or the whole thing would be pointless.