Tensor Network libraries: comparisons and future directions

Question

Can the QuTiP library be used to implement tensor network models with the same kind of features offered by the iTensor and Xerus libraries? What about TensorFlow, can it be used for this purpose to take advantage of its readiness to run on GPUs and distributed systems? iTensor seems to be better maintained, but TensorFlow is very recent and maybe it can achieve better performance (?), on top of having a simple Python interface. Are there example codes with QuTiP and/or TensorFlow implementing MPS, DMRG, MERA, PEPS, etc? Other libraries I see are TNT and OpenMPS. So many options...

I'm looking for opinions, advice and insights from users with first-hand experience in simulating tensor networks on the applications achievable by these and other software libraries around, and how they compare in performance, usability, available algorithms, and future perspectives.