As always, it depends on the situation at hand. There are several practical reasons why (HDR-)tonemapping per tile may be preferred, and there may also be positive quality consequences.
The most obvious practical benefit is that it may prove impossible to do exposure fusion or HDR assembly on multiple gigabyte sized pano files of each exposure bracket. So, to avoid having to follow different workflows for smaller versions and larger ones, doing it by tile and then stitching those seems a more logical and workable approach.
Also the quality can benefit, depending on how one shoots the tiles. Especially wide panos will have varying lighting conditions in the same scene, sometimes both front-lit and back-lit if you go wide enough. This will also result in different color balances between tiles (back-lit will have an overall cooler WB than front-lit). This can be exploited to our benefit by allowing different average exposure levels per tile (by keeping the camera on Aperture priority exposure with bracketing to span the tile's DR), and a different white balance in Raw conversion.
When one uses a liberal overlap between tiles of, say, 50%, then a proper blending engine of a decent pano stitcher can seamlessly transition between the different tile exposure/WB variations. It usually also allows to get by with fewer brackets per tile than having to cover the maximum difference found in different areas of the scene (the local DR in a tile is often smaller than the global DR of the entire scene).
Of course for relatively simple scenes, or small megapixel tiles, one may prefer to tonemap multiple panos with different exposure levels. No problem with that.
Cheers,
Bart
