Is overriding 2000 row synchronous tap query limit on XMMNewton intended behaviour?

I've been experimenting with the XMMNewton class' query_xsa_tap function, and found that any select * command would seemingly be replaced with a select TOP 2000 prior to the query being sent to XSA.

I saw that there is a limit of 2000 rows for synchronous tap queries in part of the Gaia documentation, and that anything larger should use an asynchronous query - however I then found that there is no option for asynchronously querying the XMM archive with astroquery.

Following this I experimented with different approaches, and found that using a select TOP N command (where N > 2000) actually overrode the limit and allowed the query to take place with the expected result - is this an intended (and stable) behaviour? It would be handy to know before I start using it in earnest!

Start at the XMMNewton class and its query_xsa_tap entry point, then inspect how SELECT TOP 2000 is applied before requests are sent to XSA. Compare that behavior with the XSA and Gaia TAP limits and determine whether larger synchronous queries are supported; document the stable behavior or identify the needed change.

Written by the indexing model from the issue text.