My latest experience that has brought benefits in terms of overall performance of the APP was to minimize the use of relations between GT tables and GS tables especially when used to search for data which is then used as the basis for a computed column. Or, even worse, for creating a new relation.
This completely solved the ghost lines issue I reported a few months ago
and nowadays, I reduce relations as much as possible between GTs data as well.
In other words, when possible, I replicate data in multiple tables instead of accessing them by relations.
From my point of view, this fully confirms @Darren_Murphy 's thesis, which is that the problem is not the amount of static data contained in the tables.