Madiface XT2 inputs channel order into DAW

Using a different channel map in the DAW is a very poor option. That channel map change is a hassle in some DAWs, and almost impossible in others. This suggestion just proves my point even more. For a flagship interface made by one of the most trusted companies changing either DAW channel maps or Totalmix workspaces takes time away from creative productivity and feels hacky.

This is commercial recording studio hosting many guest engineers and styles of music. Calling different sample rates the cause my problem is like saying driving up a hill is too much to ask of a car.

You do not need to use loopback at 44/48k. Seems like you have connected both converters to individual MADI inputs. At 44/48, this isn't necessary, you can daisy-chain them.
See page 17/18 of the Ferrofish manual.

One unit can combine the channels of two into 64 consecutive channels to feed one MADI input, no loopback required.

For 88/96, you would still need the individual connections due to bandwidth limitations. But to retain the channel sequence at 44/48, all you need to do is connect the second unit both to the XT (optical) and to the first converter (BNC) and create presets for both scenarios on the first converter, where either you will be sending out the individual signal from each converter to the XT or the combined signal from the first one, which will include the second converter's channels as 33-64...

I don't know if or how the converter's matrix changes at 88/96 and whether you will actually need to switch between presets or it will just work automatically. Ferrofish can help, if needed. I would assume that configuration presets also include the latency compensation setting, but I'm not entirely sure. I don't work for Ferrofish...

I'm confident that your studio engineers will have no trouble figuring this out. 

Your suggested solution would not provide consecutive channels, there would always be a gap (33-64), so this makes limited sense (even less so at 176/192). The XT is not to blame for the fact that these converters only have 32 inputs and therefore don't fully utilize the MADI connection's capabilities... If the A32 were an A64, this would not be an issue.

But that's what daisy-chaining is for...

While your car analogy doesn't really work, one might say you are blaming the car for going slower and using more gas uphill, not the hill itself... ;-)

Both Ferrofish have optical Madi ports ?
Then you can make daisy-chain with optical cabel.

You can open the housing of the connector and the two plugs split appart.
You may need an additional optical madi cabel to go from one unit to the other.

Whoops, I overlooked the fact that unlike its predecessor, the A32 Pro does not have any BNC MADI connections. :-/
In that case, you could use 4 simple Toslink cables to route all 32 channels of the second converter to the ADAT inputs of the first one, where you can route them to MADI 33-64.
Or you could get one additional module for the second converter, if you are not using the first one's MADI input.

This solution avoids having to use any Loopback and it does in fact allow you to use the third MADI input, provided you connect a device with a coaxial output. I was never referring to the BNC connection on the XT.

As I said, I am not sure how the matrix on the A32 behaves when switching between sample rate ranges, you'll have to try. You may need to save two separate presets, but that's less cumbersome than what you are doing in Totalmix and it does away with all the Loopback business.

Extremely unlikely to happen, as it would only solve an issue caused by the fact that the A32 is not an A64. Lots and lots of unused/unusable channels won't serve any purpose.
And then we'd also have to provide a solution for people using more than one Ferrofish Pulse 16MX, which only has 16 channels... Or the older RME M16 AD.

If you need to fill up a MADI connection to capacity for the kind of reasons you mention, it's best to do it externally by way of daisy-chaining.

to 1- Cool idea, Daniel. I never thought about something like this and really had to think about it for a while.
        Downside: you are wasting channels by this that you might need for other things in the future.
        And I think it adds a little complexity into the setup.

to 2- Forwarding of channels when daisy chaining / Local routing of channels on MADI devices
        It's not a "problem" once you know it. But I remember that I stumbled over it myself at the beginning.
        Once known, it is logical, but it is not really "intuitive".
        I remember, for example, the postings of two forum members, both with two Ferrofish Pulse 16 MX.

Some general thoughts about this. If this were to be solved on the driver level, you would need something like a mapping of physical ports to ports that you present to the DAW to be able to match whatever case. For MADI and for ADAT and sample rate dependent. This would add much complexity into the driver, which might even lower driver efficiency and might be difficult to maintain and also to support.
I did not hear about any manufacturer who supports this on the driver level. It might even make driver features and optimizations like in the HDSPe MADI FX driver much harder or impossible to implement.

Isn't it a waste if you double-allocate I/O ports (ADAT, MADI(BNC)) for connecting the 2nd A32 Pro
just for the sake of having channels 1-64 at double speed without a hole?
Those channels are allocated exclusively for the use case "double speed" and can't be used for other things.