scRNAseq counts change on selection

[ppavlidis]

On staging this morning:

Only 125, but then changes to 127 after selecting.

If this count is always approximate, we should indicate. Otherwise it's a bug.

[arteymix]

This is likely a cache issue.

The cache is invalidated when we update the table, but only on production. Is it also happening over there?

It works both ways if datasets are curated in staging.