Cosmic Microwave Background (CMB) observations have been used extensively to constrain key properties of neutrinos, such as their mass. However, these inferences are typically dependent on assumptions about the cosmological model, and in particular upon the distribution function of neutrinos in the early Universe. In this paper, we aim to assess the full extent to which CMB experiments are sensitive to the shape of the neutrino distribution. We demonstrate that Planck and CMB-S4-like experiments have no prospects for detecting particular features in the distribution function. Consequently, we take a general approach and marginalise completely over the form of the neutrino distribution to derive constraints on the relativistic and non-relativistic neutrino energy densities, characterised by N
eff = 3.0 ± 0.4 and ρν,0
NR < 14 eV cm-3 at 95% CL, respectively. The fact that these are the only neutrino properties that CMB data can constrain has important implications for neutrino mass limits from cosmology. Specifically, in contrast to the ΛCDM case where CMB and BAO data tightly constrain the sum of neutrinos masses to be ∑m
ν < 0.12 eV, we explicitly show that neutrino masses as large as ∑ m
ν∼ 3 eV are perfectly consistent with this data. Importantly, for this to be the case, the neutrino number density should be suitably small such that the bound on ρν,0
NR = ∑ m
ν
n
ν,0 is still satisfied. We conclude by giving an outlook on the opportunities that may arise from other complementary experimental probes, such as galaxy surveys, neutrino mass experiments and facilities designed to directly detect the cosmic neutrino background. GitHub: Parameter files for MCMC analysis and code to reproduce all plots can be found here.