XDCR v1 (goxdcr) stuck resending successfully received mutations after topology change

Please investigate issue reported on the forum that may be a bug in CAPI-based XDCR v1 (goxdcr): https://forums.couchbase.com/t/xdcr-weird-behavior/10996. It is replicating data to the Couchbase Elasticsearch Transport Plugin.

Goxdcr appears to get stuck "resyncing" after a topology change in Couchbase Server. It appears to try to resend successfully received mutations after topology change forever instead of clearing the replication queue. This is my analysis based on the user description and the mutations failed resolution stat, which according to MB-18552, indicates "that the winning revision of a document was on the destination cluster and therefore we didn't send it." I would expect such requests to eventually drop out of the replication queue.