Timestamp check currently only sets flags if the two observation values are different, and hence there's a conflict as to what the right answer was.

Merge in Rel6.1 resulted in some cases of duplicated timestamps, where the values were the same, but the metadata were different (e.g. licence policy). Although these should be fixed in Rel8, set a flag which indicates that these have been noted, and retain the most open data version of that entry.

Image from Paul Poli:

Details from Simon Noone:

The same station has been mingled into two different primary stations. See table below as example, the first station AAI0000TNCA is in Aruba and has a data policy (1) the second station USW00000411 identified as being in the US and hence (0) data policy. The cdm conversion code uses the secondary station ID (merged station id+source id = primary_station_id_2) which should be unique to join the data policy information to the cdm_lite file from the record_ID look up .csv. The record_ID.csv is created manually from the mingle list.

I think the code is seeing the two primary_station_id_2 entries and copying over two sets of all observations into the output cdm-lite file and allocated the two different data policy flags and record_numbers. I would say that station USW00000411 has the same issue.