It replaces a function call on every index lookup of a DO with a direct table lookup. That makes it faster, but also forces tables to be created that *may* not have (but most likely were), and it forces LDB to make an extra table for every single DO. I didn't make LDB that way in the first place because having a single metatable makes it trivially easy to upgrade the lib inline (you just tweak that one table and every DO gets the update), and it saves memory, at the cost of a tiny bit of performance when performing direct lookups (instead of using callback args, which you should be for the mass majority of stuff).
Long and the short of it... the performance gain is minimal, if you are seeing an impact from it, your display is probably doing something very wrong.
My question is, why the hell didn't he just submit a patch and/or find out why I designed LDB the way I did? I had my reasons not to implement it the way he did.