Inconsistent maps after loop closure, segfault at csv_export of landmarks

[MrBurnst]

Hello! I'm having trouble optimizing and exporting a merged map. I'm on the develop branch and following the multi-session use case from the wiki with sbk - aam - kfh - relax - lc -optvi on my own data of 6 individual sessions driving through some streets.

Up until pose graph relaxation everything works as expected, but after loop closure I'm faced with an inconsistent map, giving me hundreds of warnings of the sort
W0225 00:53:41.057857 3927939 check-map-consistency.cc:296] Landmark 99fa..0000 is observed twice from the same frame (42f6..0000), but the observations evaluate to [173 229] and [195 228] (distance of 22.0227) and threshold evaluates to 10

If I then try to export the map to csv files (I'm interested in the landmarks) using csv_export --csv_export_landmarks, it fails at the second mission with
F0225 00:56:22.783350 3927939 csv-export.cc:251] Check failed: it_vertex_id_to_index != vertex_id_to_index_map.end()

I tried all sorts of combinations of elq, rbl, rio, itl, kfh again before and after loop closure but to no avail. I can optimize the map with optvi but the warnings and failure remain. Any idea why that might be happening? thanks and thanks for the great toolbox!!