Thanks for all the tips, Greg. I read Rogelio Bernal Andreo's AIC 2010 presentation on making widefield mosaics using PixInsight in earnest... I like a good challenge!
The biggest challenge I've found with mosaics of very wide fields is that the field for each panel isn't flat due to the 3D-to-2D projection of the celestial sphere... usually requires panels to be either re-projected, or inaccurately registered.
Perhaps the ideal solution for perfectly registered ultra-wide field mosaics would be to follow
Nick Risinger's approach of plate solving each panel and re-projecting it to the final mosaic projection. There's a lot more work involved though...