KeyString is normally stored with the record part when used for indexes. This allows parsing code to peek at the next byte.

This is not the case for clustered collections and we may get an Overflow error when parsing strings that may contain null. We are not actually reading out of bounds but the user facing error is wrong.