Working through Seclusium, each game entity tends to draw from multiple random tables. The number of input tables can get rather large. For example, a single magical item requires around 20 tables. I think this is part of what makes using Seclusium feel somewhat cumbersome in practice, despite the large number of interesting juxtapositions.
This leads me to a hypothesis: the optimal number of tables needed to create a complex, unique game entity is around 7, probably at least 5, and almost certainly less than 10. Beyond 10, I suspect there will be diminishing returns from the the extra degrees of freedom. Less than 5 inputs, though perhaps useful for many things, seems to offer more room for elaboration (given that you are trying to create something that will play a large role and thus should have a decent amount of detail). It is probably not a coincidence that this seems related to the magic number seven.
What constitutes a single table is somewhat mutable, as two d6 tables could be combined into a single d66 table (36 entries), thus “number of tables” here is really a proxy for some other category relating to complexity, so you cannot take this “rule” too literally. That said, such combination of tables can be a technique in and of itself, as it allows the author to potentially do more work for the end referee at the cost of using slightly more storage (two d6 tables require 12 entries, while one d66 table requires 36 entries).
In general, I think this comes down more to an issue of layouts than anything else, but it is still worth considering the various tradeoffs when creating a tool. Compare the somewhat pre-integrated d100 tables of Vornheim to the many small tables required by Seclusium, for example. Another related technique is tables with multiple columns that can either be used with one roll (just read across all columns) or multiple rolls (one per column), as can also be seen in the Vornheim table below.