This looks like a numerical problem.
2 things to check:
1) that the length of pipe is long enough to capture at least one finite element of water going through it, i.e. make your timestep so small that the fluid cannot manage to pass through the pipe section within a timestep (ideally you'd like the water particle to stay in the pipe for 3 timesteps), or make your pipe artificially longer (pressure drop is not being simulated) and live with the drawbacks.
2) the engineering reference on limits to the thermal diffusivity, if any exist and try and rectify...again small timesteps may help.