I wonder if it's simply that the DTLS heartbeat is modeled on the ICMP echo-request packet? Pings also contain a blob of up to 64kB of arbitrary data (in the case of ICMP, this is/was handy for detecting pattern-dependent link problems). Unlike DTLS, though, the length of the payload is determined by the enclosing protocol, not redundantly by a length field in the packet itself.