Postgres does not assure consistency when we query a table with a filter statement. More Infos.
To avoid that, we need to apply an order by XXX on a query. Hydra does already support a primary key for every csv transformed into tables which is call __id. We will replicate that process for metrics tables.
We propose that we add to every query that we apply to api-tabular an order by __id.
Beware, when user ask to api-tabular a specific order by to process it well. For instance order by __id, COLUMN_EX