It's worth noting that the original protocol as designed by Satoshi did not have this limitation. It has evolved this way because of ad-hoc DoS fixes over time (btw I'm not saying they were the wrong thing to do, as non "ad hoc" solutions are significantly more work). But it seems like eventually a different approach to handling DoS attacks based on resource prioritisation and scheduling will become needed / implemented, and at that point the original design could be safely brought back to life.