You’re correct in that it’s a consensus protocol enforced cap on the rate of increase/decrease, and that there is no maximum enforced by the protocol.
So from a pure technical protocol PoV we can say that there is no blocksize limit, but the problem with that is that it’s generally misleading as it ignores all practical variables outside of the consensus protocol itself, such as what economic incentives miners operate under.
The reason why it’s described the way it is in the article is to (perhaps too forcefully or unprecise) combat biases often present in BTC vs ETH discussions. Specifically, there are often assumptions made on actions that BTC miners will or will not do based on rational economic incentives (e.g. not censor txs because that could get the specific miner in trouble or decrease the value of BTC) which are then not assumed for ETH miners (e.g. not vote for a block gas limit that would be detrimental to most full node operators and thus decrease value of ETH).