The one web site I saw on induction based sensors led me to believe that conductivity needs to be much higher (lower resistance) than the range you are expecting.
The diagram below is an idea which might solve your problem. It might still have problems - depends on how the ground shield is interacting with your circuit.
The idea is to prevent current flow between ground and the timing capacitor. The shield is kept at the same voltage as the cap. Current will still flow between the shield and pin 3 of the 555, and possibly from the shield to tank GND, but this should not cause a problem if your lowest resistance is indeed 10k, or possibly as low as 2k.
Edit:
It dawned on me after I posted this that you might get away with just having one probe down the center, from the capacitor, and driving the shield with pin 3.
Of course, you might not be able to put any sort of shield in your fish tank. Maybe it could be made of stainless steel fine mesh screen?