The step in [a] direction was done a few months ago with Segwit and enabling LN.
Neither of which do anything to further the goal of onboarding the world. Whether or not we have Segwit or Lightning, 1MB blocks will require 30 years to get a single transaction to each person on earth. The
only way to improve this is with larger blocks.
On the contrary, LN and similar Layer2 technology
can do much. Given a sufficiently large number of channels, kept open for a sufficiently long time, the transaction rate (summed over all open channels) can grow almost without limit. In reality, there will be practical limits on channel number and channel lifespan, so a block size increase will likely be needed at some point in the future. How big an increase, and when? I think we should find out by looking at real LN transaction flow numbers.