Neutrinos are expected to freestream (i.e. not interact with anything) since
they decouple in the early Universe at a temperature $T\sim 2\text{}\mathrm{M}\mathrm{e}\mathrm{V}$ .
However, there are many relevant particle physics scenarios that can make
neutrinos interact at $T<2\text{}\mathrm{M}\mathrm{e}\mathrm{V}$ . In this work, we take a global
perspective and aim to identify the temperature range in which neutrinos can
interact given current cosmological observations. We consider a generic set of
rates parametrizing neutrino interactions and by performing a full Planck
cosmic microwave background (CMB) analysis we find that neutrinos cannot
interact significantly for redshifts $2000\lesssim z\lesssim {10}^{5}$ , which we
refer to as the freestreaming window. We also derive a redshift dependent upper
bound on a suitably defined interaction rate ${\mathrm{\Gamma}}_{\text{nfs}}(z)$ , finding
${\mathrm{\Gamma}}_{\text{nfs}}(z)/H(z)\lesssim 1-10$ within the freestreaming window. We
show that these results are largely model independent under some broad
assumptions, and contextualize them in terms of neutrino decays, neutrino
self-interactions, neutrino annihilations, and majoron models. We provide
examples of how to use our model independent approach to obtain bounds in
specific scenarios, and demonstrate agreement with existing results. We also
investigate the reach of upcoming cosmological data finding that CMB Stage-IV
experiments can improve the bound on ${\mathrm{\Gamma}}_{\text{nfs}}(z)/H(z)$ by up to a
factor $10$ . Moreover, we comment on large-scale structure observations,
finding that the ongoing DESI survey has the potential to probe uncharted
regions of parameter space of interacting neutrinos. Finally, we point out a
peculiar scenario that has so far not been considered, and for which relatively
large interactions around recombination are still allowed by Planck data due to
some degeneracy with ${n}_{s}$ , ${A}_{s}$ and ${H}_{0}$ . This scenario can be fully tested
with CMB-S4.