Well I'm sure there is no one correct answer to this one.
For what its worth I decided to have an identical decoder in both the tender and the loco. I also fitted a SUSI sound decoder in the tender.
I thought that faffing about with small plugs is a pain sometimes so I wanted no connections. Also I felt that that the extended cab floor (is that what it is called?) that is supplied with this loco would get in the way of the connector wires and any plug I used. I think part of the thinking at the time was that if the tender could run independently of the loco then I could possibly run it with a different loco if I ever wanted too, however I've never done this.
Yes I agree that the mechanisms in both the 2015 tender and loco are identical other than the additional motion on the loco, even so I still found I had to speed match both and have slightly different setups in each decoder. I have two other sets to locos that are identical and are always double headed. Even here the setups have to be slightly different to match speeds as closely as possible, so it didn't surprise me that different setup were required with the 2015 (I think that even with one decoder two supposedly identical blocks will still run slightly differently from each other, but usually not so much different that problems are created or the speed differences are even noticed)
One of the problems with the two decoder solution is that with the relatively short wheel base and smaller number of pickups stalling/stuttering can happen at points or dirty track (although if one of the two does stop the other usually keeps going and pulls the other until it starts running again). I put a power buffer with both decoders, but that makes the solution even more expensive.
The other thing to say is that speed matching is never an exact science. A single decoder speed step may not give the granularity required, also environmental issues such as temperature, train weight (even with back EMF) etc will affect the speed a motor runs at. The direction the motor is running also has an effect. Just get the speeds matched as closely as possible, slight differences have very little effect. When doing speed matching do run the two items separately from each other.
I have a computer program that will tell me the speed a train is running at each speed step. It shuttles a train between two detectors at each speed step. Overrunning at each end allows for the locos to reach its fixed speed before the speed calculation is started.