When receiving a LS Update from the DR with the same sequence number but different data in it (such as metric) GateD needs to determine which data is newer, the arrived Update or the local database data.
According to the RFC (2328) Sequence number is the first tie breaker, in case Sequence is equal higher checksum will wins.
GateD, compares checksums in a network byte order instead of host byte order. This action causes GateD to believe that it has the newer version of the update. GateD doesn't acknowledge the DR and sends its own LS Update to the DR.
If the DR is a Cisco router , he will compare the checksums in the host byte order and find that it has the winning update.
This process will occur several times until the "Retransmission limit" on the Cisco DR router is achieved and then the Cisco router will remove the gated from its adjacencies list (perform KillNbr).
In an environment with many Gated routers involved, all of them will react the same, and won't be able to recreate the adjacencies with the DR until the originator of the update will send a newer update for this vertex.