What I would do, is take the other 1/2 of the OP amp and configure it as a buffer, but put 10K resistors in series with both inputs. (connect output to the inverting input through the 10K resistor). The point (I1), then goes through the 10K resistor to the non-inverting input.
This is an "automotive" type of environment, so some sort of transient protection is necessary as well.
Bypass caps were not shown on Mike's schematic, nor was some sort of "disabling" the other 1/2 of the IC.
If the 20mA loop supply can turn on before the LT1013 is powered, then you are blowing up the LT1013 by injecting current into the non-inverting input. Put a 4.7K resistor between R3 and the non-inverting input; that should limit the current into the non-inverting input until the other supply is turned on.
What I would do, is take the other 1/2 of the OP amp and configure it as a buffer, but put 10K resistors in series with both inputs. (connect output to the inverting input through the 10K resistor). The point (I1), then goes through the 10K resistor to the non-inverting input.
If the 20mA loop supply can turn on before the LT1013 is powered, then you are blowing up the LT1013 by injecting current into the non-inverting input. Put a 4.7K resistor between R3 and the non-inverting input; that should limit the current into the non-inverting input until the other supply is turned on.
...
If I put a resistor in the input it will change original input, isn´t it? Yes I use both halves.
I was thinking about a relay which when is energized with power supply allow the inputs, if no power then N.O. What do you think?
Can you rig the power supply wiring so that the LT1013s are powered off the same breaker that powers the senders in the tanks?
Adding the 4.7K resistor between the top of R3 and the Non-Inverting Input (nii) of both LT1013s will have no effect on the operation of the original circuit when it is operating normally, actually it will slightly improve it. The input bias current of the LT1013 is ~30nA, so that during normal operation, the voltage drop across the 4.7K resistor is only IR = 4700*30e-9 = 141uV, which slightly reduces the existing offset voltage (a good thing).
If the loop supply is on, but the LT1013's supply is off, the 4.7K resistor will limit the current that would flow into the nii to less than 4mA, preventing blowing the nii on the LT1013...
The data sheet says +-22V, so up to 44V from V+ to V- pins. In this circuit, we connect V- to 0V, so 28V (or 31V with the battery bank during charge) is ok.