Such a long-term regression of network difficulty means nothing else that transaction revenue from active users is not able to fund current network security anymore for both: active and passive users. Delaying of halving is simply: not introducing additional more damage to the network security. as a side note: it's really a conservative approach then - and would fit to Bitcoin, in fact.
And obviously we are approaching post-subsidy era gradual way, so this so simple mechanism above will trigger in the moment of passing equilibrium described above. And effectively we will stay in this equilibrium, having simultaneously:
  • as low on-chain fees as possible
  • as low subsidy as possible
...while keeping the network security robust and all that balanced by the free market (in its finest) between these two competing parties.