I think I can answer a few of your questions.

> I see that the HDSPe MADI FX can do TotalMix FX with DSP Effects. Do I really get full EQ, Dynamics,
> etc on every channel the same way I would on a console? Are there limits here? And is there processing on outputs?

Unfortunately, I can't say exactly how many FX effects you can activate in total. But allow me the following observation.
Many years ago, I tried this on a UFX. The result was that you cannot activate FX on all channels. The design certainly includes something like a "mixing calculation". It is very unlikely that anyone needs full FX facilities on all channels. Most of it is done in the DAW or in combination with VSTs anyway. I use a lot of the TM FX functions, but I have only needed EQ in special situations, such as in monitor or headphone mixes, but not on every input. After all, the goal when recording is to record the signals as cleanly and unaltered as possible. So why include an "FX Power2" in devices that would only cause unnecessary costs for almost all users when they buy them?

> Can you set up DCAs like a console?

You could e.g. create Fader groups, AFAIK 4 and a temporary. A DAW like Cubase might be more flexible in that regards allowing for more groups.

> What kind of performance can I expect? And what about latency?

Could you kindly be more concrete what you regard as "performance" here ?

You are also very unspecific, which latency you mean. On the card itself, or do you mean round trip latency ?
For what device ? At what sample rate ? With what ASIO buffer settings ?

I can tell you generally, that RME products have excellent drivers and very low latencies. No matter whether on the device itself (near zero latency monitoring ) or through USB or PCIe.
See here the RTL of different RME products which I had access to.

> Can the cue bus be routed anywhere?

You have a low latency CUE and PFL function in TM FX. You can assign it to one of the outputs in the TM FX control room (Main Out, Phones 1 ... Phones 4) using the assign button.
Inputs and SW Playback channels can be set to have either Solo or PFL function via TM FX -> Options -> Settings.
HW Output can use Cue.

> EDIT2: UFX+ won't work since it doesn't appear to allow both MADI ports at the same time.

According to the manual the WC ports can be used for MADI coax in two modes
- coax being a mirror of optical MADI port
- split 32ch optical / 32ch coax

But at the end of the day the UFX+ has only one MADI bus. This allows for 64ch at single speed and 32ch at double speed.
The question is whether this is sufficient or not.

> While I only need a stereo mix for output for program, I could use some analog inputs. I would only be using
> two optical MADI ports, so I could use an interface on the coax MADI. Can anyone recommend something that
> has just a few analog inputs and outputs on a single device?

Do you mean a recording interface or an ADAT or MADI based external device ?
It would be helpful if you could sum up your minimum requirements.
What sample rates are you using ? Single, Double, Quad ?
How many MADI channels do you need of what type ?
How many analog inputs and outputs do you need ? On the device itself or via ADAT/MADI extension ?

> OR, can I put a 2nd RME card on the same system to get more analog I/O?
> I know that in DAW software you have to select one ASIO driver, so I'm not sure how that would work.

1st of all, when using multiple cards then you need to synchronize them either by WC or by using 1 digital connection between them, e.g. on the clock master one adat TOSLIN to the next cards input and this card can read then the clock from this input. You need manually ensure in the RME driver settings, that all cards are using the same ASIO buffer size.

You will get a TotalMix FX instance per recording interface and on TotalMix FX level you can not route channels between those interfaces only inside of such an instance.
If you perform routing on DAW Level then this is no issue, because if you use two or three cards of the same kind, then you can access them all through the same ASIO driver.

Caveat: RME drivers usually support up to 3 cards of the same type, but there can be certain limitations on the computer side whether the system can work with so many cards and channels and this also depends on the sample rates being used.

I think one of the best cards is the HDSPe MADI FX, because
- its the only PCIe card with full implmentation of FX chip
- the card has a unique performance optimizing driver, groups all channels in groups of 8 and if no channel in such a group is being used, then the driver does not allocate resources for all of these 8 channels. So if you organize the connections properly and use maybe only half of them, then only 50% of the channels are being transferred instantly over PCIe not the whole amount of channels. And if you only use 2 channels in a project, then the resource consumption is only minimal. This is one of the best performance saving drivers.
- it has additional analog phones outputs, so that you can directly monitor using phones (very nice and flexible)
- it has AES I/O You could e.g. very nicely connect an ADI-2 Pro R BE to have a very nice phones preamps as well as best features for monitors and additionally also high quality analog inputs.
- if offers 3 MADI busses, 2 optical, 1 coaxial and the coaxial can also be replaced by a different daughter card so that you can have even three optical MADI busses
- offers MADI routing and Redundancy features

You can look here at my review of the HDSPe MADI FX:
https://www.tonstudio-forum.de/blog/ind … Pro-FS-BE/
The full review in PDF format:
https://www.tonstudio-forum.de/index.ph … -v1-1-pdf/

Interesting thought. I'm familiar with AVB. The switches required for it might make it expensive. Dante would be easier to do but not having FX would defeat the purpose of this sort of application.