Currently all orthophotos are orthorectified. The process of generating such assets however involves several steps (structure from motion, meshing and texturing) which tend to be slow. Sometimes users might want to simply stitch images together to get a rough overview of the map (this is a frequent use case in agriculture). It would be interesting to explore methods and implementations that do fast image stitching, while keeping georeferencing information from available GPS information.
This is the approach that Pix4DFields takes.
Obviously a viable approach.
Incidentally, I did make use of it a number of times when fast turn-around was paramount for the client compared to more precise results obtained from my regular Pix4DMapper workflow.