Thanks to Ed Felten, Frederico Lacs and Patrick McCorry for comments and feedback
Blockchains have failed to meet the throughput demanded of them. Since the end of 2016 on Bitcoin and the end of 2017 on Ethereum the frequency of full blocks has shown that demand for blockspace is outstripping supply. Transaction fees have risen, pushing out the users and use cases that can’t afford them.
The naive response is to increase the supply - bigger blocks - but this has knock on effects to the security and decentralisation of the protocol. Bigger blocks means the nodes in the network need better hardware to validate blocks. Node operators will need to pay higher costs to purchase the hardware, meaning less of them will do so. But to remain secure a wide variety of users need to be able to validate blocks.
Instead the search is on to find methods of scaling that do not increase the hardware costs of the average user. However some methods of increasing throughput are one-off improvements - eg a more efficent data structure for storing state data, or a cheaper method to verify user signatures. In these cases the gains will not last, increasing blockchain demand will eventually consume a one-off increase in supply. At that point another improvement would need to be found to once again provide more supply.