The conclusion on duplicate delivery at the end of the paragraph above NOTE 1 is wrong. Section "7.4.3 Base recovery function" describes the relationship of the (1) size of SequenceHistory variable and (2) the difference in the number of packets in flight on all paths. It is fine, if the "receive window" covered by SequenceHistory is too small, then packets on the long path are not considered (discarded) by the elimination function. Later in section "7.4.3.4 VectorRecoveryAlgorithm" such discards are stated explicitly: ... "Rogue" packets, meaning packets outside the frerSeqRcvyHistoryLength window, are discarded as invalid ...
However, the text in 7.4.3 makes a wrong conclusion as it says that duplicate delivery occurs. It is stated both in NOTE 1 in 7.4.3 and in the paragraph right above NOTE 1.