By using the floating point limiter that you have drawn does that take U8 out of the circuit?

Yes.

The 100R trimmer that could be a fixed value once tests are done correct?

No, unfortunately not. FETs can show drastical tolerances and a fixed value would make problems very probably. A bipolar current limiter is less sensitive to manufacturing tolerances. But current limiters with bipolar transistors often suffer from higher voltage drop.

Is the heatsink necessary for Q2?

Calculate it by yourself, whether Q2 will become too hot.

What do you think about making the circuit isolated with a 1000V isolated between ground in this circuit and the ground the customer hooks up on pin 10? Of course that would add a lot of circuitry, space and cost but then everything would be floating.

If this means, that current limiting circuitry arround Q2 and U8 remains unchanged, then it's not a good idea. Why? Because U8 is always in danger, that with fast voltage step at inverting input it will not be able to shut-down Q2 fast enough. In any case I would use a real floating current limiter.
Isolating your circuit is one method to solve the problems arising from high common mode voltage. Another is to use real instrumentation amplifier design. Have a look at datasheet of INA117 and SSM-2402, to see how very high common mode input voltages can be handled. But in any case, if you really want to achieve 22bit precision (I noticed the ADS1212 in your schematic), then unavoidable common mode impedance will make you trouble. Especially, if you connect two or more recievers in series. I don't think that more than 12bit resolution can be achieved then.

If you proceed with isolation method, keep in mind, that all kinds of interference, ESD and common mode noise will hit the barrier. So, a varistor or something similar, which allows to pass very high overvoltages from either ground to the other, will help to make your product living longer.

Kai