Hi Latif,
Thanks for the detailed information including the yt version and the output!
What you're experiencing here looks like normal behavior actually, at least in my experience. When the merger tree tracks halos back in time, it goes from timestep output to earlier timestep output, looking for the halo at a previous timestep which contains the largest number of the particles in the current timestep. Unfortunately, at early times, like z=30, very few halos have actually formed because they just haven't had enough time for gravitational collapse to take over. Thus, the merger tree fails because it cannot find a parent halo at that timestep.
However, it might be worthwhile for us to make the error message more explanatory so that it's obvious what is happening. I'll make an issue for this to be corrected in the future.
Cameron