KCharSelect initially started out only supporting the basic plane U+xxxx. Support for code points U+1Fxxx was added mainly for emoticons support (remapping them to the private code points at U+Exxx), but there are many other blocks in supplemental plane that cannot be added by remapping.
The kcharselect-data file needs to be extended to support code points larger than 16 bits. Additionally, CJK script also uses U+2xxxx. With more than 80K code points, we don't want to add them as a big block to the table, but want to support selecting by radical and browsing by stroke count. We could also incorporate UCSUR and math related data.
Because of these changes, it makes sense to completely redesign the format for the kcharselect-data file, splitting it into multiple files for the different purposes.