Adaptive schedules, i.e. those where block size limit depends not only on block height, but on other parameters as well, are surely attractive in the sense that the system can adapt to the actual use, but they also open a possibility of a manipulation.

E.g. one of mining companies might try to bankrupt other companies by making mining non-profitable. To do that they will accept transactions with ridiculously low fees (e.g. 1 satoshi per transaction). Of course, they will suffer losees themselves, but the they might be able to survive that if they have access to financial resources. (E.g. companies backed by banks and such will have an advantage).

Once competitors close down their mining operations, they can drive fees upwards.

So if you don't want to open room for manipulation (which is very hard to analyze), it is better to have a block size hard limit which depends only on block height.

On top of that there might be a soft limit which is enforced by the majority of miners.