Electrical – 4-20mA loop voltage to high

1. Shorting. As Olin Lathrop mentions, the power supply (at the receiver) sending power to the sensors should be current-limited. Say the power supply driving the loop is +40 VDC, but current limited to around 30 mA -- perhaps by using a LM317 and a 47 ohm resistor in the "current limit" circuit shown in the LM317 datasheet and the " Simple circuit to limit current through a FET " question. (It doesn't matter if the current limit is in series with the "hi" line going out to the sensor, or in series with the "lo" line coming back from the sensor to the 250 Ohm resistor, or both). If something horrible happens and the two lines going out to the sensor are shorted, the LM317 limits the current to 30 mA through the short and through the 250 Ohm sense resistor, resulting in 7.5 V across the sense resistor (or perhaps only 5.6 V across the resistor with a different configuration of ADC-protection diodes), zero V across the short, and the rest of the +40 VDC is across the LM317 current limiter. That gives at worst 30 mA * 7.5 V = 0.225 W, so a single common quarter-watt resistor can (just barely) handle it indefinitely. The LM317 has thermal overload protection and can withstand the full 40 VDC and 30 mA current indefinitely.

2. Isolation. The two wires running out to each sensor should be insulated, and each sensor itself should not have any other current-carrying wires connected to it, so each sensor is fairly well isolated from all the others (even if the "hi" lines for all the sensors are connected together at the receiver). The current going into any well-insulated object should be equal to the current coming out of that object -- electrons generally don't jump out of wires -- so I don't see any path for "ground loops" anywhere. The main advantage of current loop sensing is that it's much easier to insulate current-loop wires to give practically zero leakage current than it is to make voltage-sense wires highly conductive to give practically zero voltage drop.

3. Multiple receivers. No matter what you do, you can't have hundreds of receivers along a loop. I suppose you could sacrifice some noise rejection and use 50 Ohm sense resistors to get a few more receivers before it doesn't work at all. But eventually, you're forced to either (a) use "Current Loop Repeater" devices, or else (b) use "zero voltage drop" current-to-voltage converters aka "transimpedance amplifiers". Alas, some receivers assume that they can connect one end of the 250 Ohm sense resistor to safety ground -- that works fine as long as that's the only receiver that does it; connecting merely 2 such receivers in a loop to a sensor causes bad data.

4. Surge suppression. Sounds fine, possibly already a bit of overkill. Perhaps add a Zener to protect the ADC from over-voltage and reverse-voltage.

I don't recommend you to use the design in the picture. OAs ( Operational Amplifier) are good for Voltage gain and High impedance output, but not to "play" with current, the current output is sensible with the components tolerance. Try to simulate your circuits with variation of 5% in your devices (resistors & VCC) and see the output. When you will buy the components they will have deviations about 1% to 5%, depend on the quality of the materials you will buy.

Transistors are the devices to play with current gains, but impedance could be the problem.

See also the tolerance for the Valve mid-position for current, it is 1.3%:

Voltage type [V DC] 5 (±0.1)
Current type [mA] 12 (±0.16)

Your are right the 12.267 mA are not correct ( max. acceptable would be 12.16 mA), take this into consideration because it means: the tolerance current output of your circuit must be max. 1.3%!!!.

It could means also that maybe temperature compensation of your output must be implemented depend on the range of temperature where the device will operate and ICs you will use.

I have seen the Page 6 Datasheet of the valve:

Setpoint value
Voltage type [V DC] 0 … 10
Current type [mA] 4 … 20

Why don't you try to control the valve with "Voltage type"? I think it is easier to transform your +-5V to a 0-10 Volt linear signal with mid-position on 5V. And you have a little more tolerance (2%) requirement for your output. It is easy, you can use an OA adder with a two inputs: one a +5V offset and the second your +-5V signal (you must check the tolerance).