If you want to use DISTINCT this is exactly the way to go, DISTINCT is executed in the return step so you need to save the distinct set of vertices before executing the VISITOR on all of them.

If you do not want to use DISTINCT you can also write:
FOR v IN OUTBOUND @start @@edge FILTER … RETURN VISITOR(v)
which will be more efficient

About the visitor function, just to make it clear, are you saying that one would e.g. have code like:
let vertices = (AQL using outbound/inbound/any with optional DISTINCT and filter etc)
Then just run a regular for v in vertices VISITOR(v)

something like that?

Hi, the intention of the new traversal syntax is to be more simple and use less options than the function-style traversal. This simplification allows us internally to use several shortcuts and optimizations and allows for an easier entry point for a user. Also we think they cover a lot of use-cases already.

Therefore there not (yet) any plan for further options except the direction and the steps for these traversals. The “visitor” function should be implemented in the later AQL statements (which covers every visitor just returning or counting attributes). Duplicates can be removed using the DISTINCT modifier.

However if these features are not powerful enough you can still use the function-style traversals which are indeed more generic and powerful, but also more complicated to configure.