This paper evaluates four forwarding strategies for fragmented datagrams in the Internet of Things (IoT). We focus on classic end-to-end fragmentation, hop-wise reassembly, a minimal approach to direct forwarding of fragments, and direct forwarding utilizing selective fragment recovery. To fully analyze the potentials of selective fragment recovery, we include four common congestion control mechanisms. We compare all fragmentation strategies comprehensively in extensive experiments to assess reliability, endto- end latency, and memory consumption on top of IEEE 802.15.4 and its common CSMA/CA MAC implementation. Our key findings include three takeaways. First, direct fragment forwarding should be deployed with care since higher packet transmission rates on the link layer can significantly reduce reliability, which can even further increase end-to-end latency because of highly increased link layer retransmissions. Second, selective fragment recovery can mitigate the problems underneath. Third, congestion control for selective fragment recovery should be chosen such that small congestion windows grow together with fragment pacing. In case of fewer fragments per datagram, pacing is less of a concern but the congestion window is limited by an upper bound.