My QGIS project has various raster and (complex) vector layers from different sources, in different CRSes. Using reprojection on the fly, I don't have to reproject them. But should I, especially for the layers I've decided are part of my base map set, for performance or other reasons?
In particular, is the performance overhead associated with ongoing on-the-fly reprojection of a shading (DEM-derived) raster layer, or a complex and dense vector layer such as contours or vectorized waterbodies, worth manually reprojecting and saving in the displayed CRS?
Editing to add: my goal is a usable, responsive map that will continue to be improved with adding new information, rather than analysis of embedded GIS data.