He says very little of recovering from the majority hashpower attack.

Lets assume tomorrow XMR suffers from the majority hashpower attack as the result of government control of the four largest pools. The attacker has now 90% of hashpower
and is now mining 90% of blocks (mostly empty blocks, the only transactions included are some whitemarket KYC-ed transactions and the ones of selling for fiat 0.6 xmr of coinbase reward from the tail emission in order to recoup the cost of electricity, the hardware atc.).
The pool of unconfirmed transaction rises with fees incentivizing the black market miners, and they are now able to mine every one in ten blocks. But under the dynamic blocksize rule set they are likely to loose the 0.6xmr of the base reward because their blocks are going to be much larger than the empy blocks mined by the hostile miner.

Let's divide mining into three types:
1. Miners reward = base reward + fees (Bitcoin now)
2. Miners reward = conditional on the blocksize base reward + fees (XMR now)
3. Miners reward comes entirely from fees. (Kaspa after 2037 and BTC in the very distant future).

It should be the easiest to recover from the majority hashpower attack in third type. Cases 1. and 2. are more difficult.

P.S. What gives me hope is the culture of XMR and Kaspa when compared with BTC that is conducive to hard-forking upgrades. When under attack
XMR may hard-fork and strip the reward in order to overpower the censor. BTC is very unlikely to do this.