Water meter sensor

For nearly 5 years I've been monitoring water usage.  Thoughtfully, the water company provided a meter with a special slot to insert a reed sensor, which I used for about a month until they decided to install an automatic reader. Whilst having brass case, there must be some magnetic shielding, as a reed sensor adjacent detects nothing.  
Undeterred I bought an AH3503 electronic sensor. A single one will detect the meter magnet, but the variance is in the order of 30mV. Detectable with a comparitor, but it doesnt take much to ruin the reading. So I put 2 sensors back to back into a differential amplifier. It worked well, and I got an output of about 0.6V to 2V with the meter. I made a sensor, and put a ferrite in for good measure, as this somehow magnified the effect.
Several times per year I had to repair it, usually some sort of water ingress (the meter is underground about 70m from the house) despite conformal coating. You can see the home made vero board and the black heatshrink (~65mm long) removed from the sensor pair on the right. The adjacent device is the sensor and this was the latest failure. Despite being sealed in glue lined heatshrink, verdigris got in rotted off the legs. I started off with a LF351 in a socket, then removed the socket, then bodged in a LMV321 rail to rail device covered in wax.

new pcb left, old sensor + circuit right

new sensor sealed up

I had a bit of space left on a recent pcb order, so I made a compact replacement sensor. It's small enough (double sided) to hold both sensors soldered on the bottom, 4 resistors, 2 capacitors, a diode and a SOT23 LMV321 opamp. A very respectable output of 0.06V to 4V (5V supply). The finished device (4 coats of hair spray) and heatshrink is little bigger than the original sensors alone. The output in situ is 0.1V to 1.6V as the magnet moves in the meter. Supply volts are around 4.6V once it's made the 70m trip on cat 5 CCA cable. Current consumption is trivial.
Hopefully the water company won't think I'm trying to frig their meter somehow. I say their meter. I had to buy it and pay for the install.
1/6/20 random readings.  Either the sensor has been touched or it just moved. It was at an angle so a bit far away to register. Tested with tip of little yellow screwdriver and got the range as above. It was below the official sensor, but it fits above a lot tighter. In use it goes from 2.215 to .