If I understand you correctly, you need to add a little damp to the fluid, as well as some density dissipation. The fluid is dynamic, where each state depends on the previous state, so you need to avoid things like velocity and density accumulation so that after running the simulation a while( with looping input emission ) that it settles into a repeating loop. Also if there is turbulence(with non-zero turbulence speed) then it could break any looping. The turbulence on the fluid does loop, but offhand I can't tell you the turbulence speed you would need to loop over a given frame range. You could find out for yourself by just running a fluid with high damp, velocity display and turbulence, then look for the loop point.
Or you could just turn off turbulence if you don't need it. There is also turbulence on the emitter, which will also loop in a simular fashion.
At any rate the basic technique is to loop all inputs to the simulation then run the simulation for several loops until it stabilizes then set the initial state. The damp needs to be large enough that the simulation will stabilize, and the dissipation needs to be great enough that density stops building up after a while.
Duncan