Sum/Difference Op-Amp circuit drawbacks that no one mentions?

[Terry Bites]

They are aways a disappointment. If you dont need speed or high impedance inputs the INA105 ( INA133 INA157 etc) are more accurate than a hand rolled.
the DC and AC errors for the inverting and non-inverting inputs are not the same- welcome to hell. There is always a trade off between DC accuracy and speed. Her's a good article on error analysis in diff amps: https://www.dataforth.com/catalog/pdf/an107.pdf 

You generally can substitute instrumentation amps for diff amps. It will be more accrate from the outset and it can work out cheaper than the cost of a precision opamp and a few precison reistors. Less PCB room as well. The AD8421 and the LT1102 are fast. Check out the OPA396 and INA849 to get an idea of whats out there.

[LoveLaika]

The more I work on this, the more I'm running into issues: loop gain, possible crosstalk, phase shift, bandwidth, etc. Running simple simulations is making me rethink this whole idea. I kind of need high speed, trying to achieve a bandwidth of 100 MHz.

Well, I suppose that this is a 'difference' op-amp configuration, as in (0-{INPUTS}) if you look at it the right way. I've been testing with op-amps like the OPA818, but I never really thought about using other types of op-amps. How do instrumentation op-amps handle multiple inputs, as with my multiple inputs shown at the non-inverting terminal?

[TimFox]

Do you need DC to 100 MHz?
If only AC (RF), you could consider using broad-band transformers (such as the Mini-Circuits Labs units) and low-value resistors (say, 100 ohms) to do combinations of inputs, while respecting transmission line impedances.
Anyway, if you use operational amplifiers up to 100 MHz signal frequencies, how will you verify the accuracy of the result, and to what tolerance?

[LoveLaika]

The only way I'm verifying is using my signal generator to output a square wave and comparing the expected output on an oscilloscope. Caveat, my signal generator doesn't go up to 100 MHz, but it does go high enough where it is considered acceptable. And, yes, I'm looking at DC to 100 MHz or close to it.

Would a current feedback op amp be good here? If the bandwidth isn't tied to the gain, I won't have to worry about how I need higher gain when there are more inputs.