ROME/REA timeseries.  Our contact is Yiannis Tsapras.

Oct 2024: upload of a full data release via rsync.  Then nothing
happened for a long time.  Resumed work in Jan 2026.


Test set
--------

The problem is that the first many lightcurves are all NULLs (in
general, the ratio of data/NULL is *very* odd in this data set).
We therefore cut out some objects from the crossmatch fits and the
matching hdf5, too.  This needs custom code.  Run

  python res/make-sample.py

in the resdir of the full thing and copy the resulting files into the
test system's data.  This then means that the ids are all wrong, and
we have a hack in place in get_photpoints that maps ids 70001f to 1f if
70001f aren't accessible.  Let's hope this never explodes.  Oh my.