All device must advertise using the 1Mbps PHY for compatibility, devices that can use the 2Mbps PHY can use the 1M to jump to 2M.
You can not have GATT communication on 2M if you didn't advertise on it.
MTU affects whether you need to do long write while data length affect how efficient the lowest level send out packet, a big MTU can be inefficient if you have to split it too much although big MTU will giver upper layer better ergonomics
While ATT is the most commonly used protocol to talk, other high level protocols such as CoC can be used that improve throughput for other applications
Extended advertising may not be implemented even though devices implement BLE 5