This is a list of Jan's guesses what parallax solutions in Gaia
eDR3 are actually spurious.  Data was rsynced up to us by Jan.

Test data: healpix 3663 (lvl5_res_snr45_both_3663.fits) in accordance
with gedr3dist.

Dump creation:

psql gavo -c '\copy (select * from gedr3spur.main order by source_id) to stdout' > static/dump.txt
pigz static/dump.txt

Sample creation based on that:

zcat static/dump.txt.gz | head -3000 | gzip > sample.txt.gz

Yes, this RD reads from sample/* rather than data/*.  data/ contained
the sources for the intermediate tables as of ~svn rev. 8230.


How I built the data
====================

The data as in the dump was assembled on-site.  I originally got from
Jan:

* lvl5_res_snr45_both_*.fits containing the version 1 fidelities
* lvl5_*.npy containing version 2 fidelities as numpy dumps
* neighbor_table_*.fits.gz containing neighbour distances and some
  derived quantities

From the npy files I made fits files using a small python script so I
can simply use the booster infrastructure.

The fidelities I pulled into two tables, gedr3spur.v1 and gedrspur.v2.
The RD that did that is still in the SVN history.