Truth to be told, A* over navmeshes and path smoothing ('string pulling' by eliminating points with LoS) with adding occupancy/weight to individual triangles (or polygons? Not sure what's used here) is nothing particularly new... going some 25 years maybe?
I've not been in triple-A games for many years, but I'm not sure I'd choose navmeshes today for an RTS style game over a large terrain - they tend to be static in nature (without attaching a lot of additional data to them, projecting dynamic obstacles onto them, re-weighting all the polys all the time, while making sure that the resolution is fine enough for this purpose or coming up with some vector field on top of the larger ones.... but no one needs to save RAM these days, do they... it's not 32MB of Playstation 2 anymore).
Edit: forgot to add that once you realise that A* is 'open ended' and that you drive it via the heuristic estimate, that can be anything, it becomes much much more suited to 'strategic' path-finding than what's normally envisaged described as 'A-to-B' path finding.
I've not been in triple-A games for many years, but I'm not sure I'd choose navmeshes today for an RTS style game over a large terrain - they tend to be static in nature (without attaching a lot of additional data to them, projecting dynamic obstacles onto them, re-weighting all the polys all the time, while making sure that the resolution is fine enough for this purpose or coming up with some vector field on top of the larger ones.... but no one needs to save RAM these days, do they... it's not 32MB of Playstation 2 anymore).
Edit: forgot to add that once you realise that A* is 'open ended' and that you drive it via the heuristic estimate, that can be anything, it becomes much much more suited to 'strategic' path-finding than what's normally envisaged described as 'A-to-B' path finding.