Reparameterizing gamma distribution to avoid low BFMI

As I am the one who came up with this idea, I would love to know how using this reparameterization helps your situation.

In a nutshell, reparameterize the distribution in terms of its mean and its shape parameter by writing your own _lpdf function like was done in a reply to my link. In the short example there, this method had better sampling than the built in gamma distribution. Good luck!