If you apply basic filtering to that node count (nodes that did not update their state in the last 60 days, nodes with no advertized adddress, channels disabled on both sides,,..) You already diminish to 3500 nodes / 16500 channels /1800BTC currently, according to my node data.

So it would be interesting to see the historic data with such filters included. Because if the number of actually useable nodes is still 1/4 of the raw advertized number, i am not sure we can actually draw advanced conclusions from it