So far I've not found it to be particularly useful.
Just tried York to Brighton now.
Both are routeing points so that's typically valid (very broadly speaking ) ...
Via London: via ECML direct, or via ECML+Cambridge , or via MML; (and then BML) [AY EY KY]
Avoiding London via MML to Birmingham/Coventry Oxford -> Gatwick etc. [XB+BY]
( With, of course, a myriad of of possible routes within these broad descriptions.)
But on NRE you get back maps of four "valid" routes - each of which is identical at the resolution presented ( all ECML direct ->LONDON->BML). If you zoom in and look very carefully the four routes are identical apart from their slightly different routes through South London.
When I first thought took a look at it I concluded that the problem is they're just presenting the first n answers on a depth-first search ( so you get trivially different answers that are clustered around origin or destination.) I guess that could still be true if the planner is splitting the search at London.
There are potentially far too many different useful routes to display in this route-by-route format : the best way to show other valid routes would show the maps instead (consolidated and suitably redacted and annotated for easements/doublebacks.) That's not a straightforward thing to do.
If they're constrained to showing n individual routes then they should have need a measure of difference in the presented results so you at least begin to explore the space in a useful way (rather than showing almost identical answers.)