PS for the Transform Grid tweaking to hand tweak alignment of images, sometimes to fit the frame.
PS for spot removal, as LR keeps it as instruction and can slow things down.
PS for "knocking out"/masking.
I like Neil's explanation of how the creative is in LR and the technical in PS.
I agree, although softimes I end up giving some color intensity in PS.
The pixel level adjustment is something I really need. I also like the HPass filter.
Also the fact that they are differnt, as Andrew and others point out, Pixel vs Instruction. You can "feel" this while working.
PS for detailed DodgeBurnSat (not too often).
Most all images are in LR first, and get worked on first, then Off to PS for some final pixel level touches. You can say 80-95% of the image is done in LR.
I also like how the response is in PS. If I make adjustments, they are accurate and instant.
LR feels heavy. This is likely my catalog size or disk access? (my hardware is OK).