+1 on the comments below by Thomas.
"Fee market" is not a
binary option, either on or off. Like all markets it exists in varying
degrees over time and with more or less influence on the process of which it is
part of. As it stands now, and likely for another decade at least, the
fee per tx constitutes a very, very weak market signal for the miners since it
makes up less than 0.5% of the rewards paid for mining a full block. This is the normal and expected scenario, aimed at driving use of Bitcoin
as widely and cheaply as possible so that network effects will cement
Bitcoin for the future, for all users.
As a comparison, how many people would have
taken up using email if there had been a per-message fee and hourly
sending limits with only the highest-fee messages being delivered? And
how would it look if the email protocol developers imposing fees
& hourly sending limits were pushing hard for use of another email-delivery protocol on
top of the existing one?
How would email have looked if it required 300 MW of power to support it for "free" for 10 years?
In practice email was never free- it was paid for by the payments users made to ISPs. ISPs paid for email and network infrastructure from that.
The equivalent analogy here would be to drop fees completely and pay a specific miner to mine all of your transactions as a monthly subscription (which of course doesn't work in a non-permission-based network).
Miners have real (huge) costs - they will be in a lot of pain with reward halving if a few model does not replace that. That in turn poses a huge risk of smaller miners shutting down, which in turn centralises things even more. I would argue that the lack of pool diversity and thus lack of block makers is already the single biggest risk for a decentralised system; avoiding the issue of fees just accelerates this.