Oh fun, I can speak to this. I was at Janelia for the last year and a half, and actually just submitted a paper with Keller. I did ML stuff for reconstruction and otherwise post-magics with the images.

There are a lot more people in that place working on scopes than just Keller and Betzig! Whole institution is basically vertically integrated fluorescence microscopy. You got your folks that work on making specific parts of mice or fly brains glow, your folks that build the special scopes, VR rigs for the mice to run around in while they're getting imaged...

IMO we are close the limits of what can be done with 'classical' microscopy, classical meaning basically as practiced now. The real strides are going to be made in sophisticated reconstruction techniques leveraging the latest and greatest in ML and high fidelity simulations of the scope itself, which allow us to get close to reconstructing at the physical limit of information captured by the scope.

But even with current 'classical' microscopy such as Keller et. al's Simview, the datasets we are collecting far exceed in complexity the neuro communities ability to analyze them, in particular at Janelia. Amongst other frustrations I have with the institution is the complete lack of interest in serious theory. Instead its all grasping at straws with very expensive and sophisticated straw grabbing apparati.

Weird place, amazing science nonetheless. But my advice for neuro people considering a visit is to stay far away.