Hi Yoram,

As Ralph points out, even in 6.0.3 we do not recommend module sizes above 10k and in 5.0.2 it was the same.  If you create 20k artifact modules you have to appreciate that you're introducing a trade off sluggish fetch times for the users.  This needs to be strongly evaluated (larger queries hitting the server; more focus on one area of data etc) before you continue down that path.

The problem here though is likely to be your access times to the Jena indices.  So see the chart here for knowing where you are on the overall scale.  1 user and 1m artifacts is still going to need some decent hardware.
Also note that in 5.0.2 the indices were larger.  Up to a third in fact due to the comments, so where you indices reside is going to be hugely important.  SSD will make a night and day difference to your system.

A word of warning on manual refresh.  You are undoing all the caching and causing a refresh from the server, so once you know that you realise that this is an expensive operation and you probably want to avoid it.

If you look at a .har file output (or similar, Developer tools in Chrome) then you can see if the delay is network related or purely fetch times from the server.  The server itself is also going to report out in the rm.log if it is being overworked.
You might also encounter database delays, so check that for latency & ensure that server is not under strain...although the behaviour you describe is more likely indices.

Further than that you may need to open a PMR with the Support team to get to a root cause, but if you exceed 10k artifacts in a module...you know what we're going to say first!!

Finally, the Performance report for RDNG 502 should give you an indication of what timings to expect as normal for the hardware you have.

Kind regards,
Paul