In human cochlea tonotopical organisation is that higher frequency sounds are detected near the entrance whereas lower frequency sounds are deep inside the cochlea. Is this tonotopical organisation (high frequencies at the entrance, low - deep inside) justified by some physical or biological constraints or it doesn't matter and we could in principle have cochlea where low frequency sounds are detected at the entrance?
There are indeed constraints: in principle, low-frequency sound waves travel better in viscous media (look here for a simple explanation). The idea behind that is that being sound a vibration in the medium, it causes its molecules to move. The faster the movement (that is, the higher the frequency) induced by the wave, the more the particles will disperse their energy by crashing into each other. This leads to more loss of energy for high-frequency waves compared to low-frequency ones when travelling in a medium.
Now, all of this leads to the detection of high frequencies being better at the beginning of the canal. However, here is the law that quantifies the loss of amplitude of the wave with distance. As you can see, the damping is higher for high frequencies, as well as high viscosities. However, plugging in some sensible numbers (20 kHz for very high-frequencies heard by humans, 1500 m/s for the speed of sound in endolymph, 1 centipoise for its dynamic viscosity, 1 kg/liter for density) gives a negligible loss of intensity (something like 0.0001. I'm sorry for the poor formatting, but I'm on my phone right now and I hope you will try to run the calculation yourself).
Now, a possible need for the accurate detection of even higher frequency sound waves in water for the fish in which the cochlea appeared is possible. However, I neither know if that is the case, nor if the difference would be a significant one.