I don’t think it really matters, if you follow the Fair Access Policy you will probably never send messages directly after each other.
If you do need to send multiple messages right after each other a per channel implementation will be faster. Since you can send on the 2nd while the first is still waiting. But that will only be faster if the number of messages you want to send is equal or lower to the number of channels in the band.
Else, if you send a message every few minutes, the only difference it will give is the implementation you have to do. Thus it would be your own judgement which is easier to implement. (If it is not already implemented by the chip/library you use.)