Though I am not a specialist of kinetics, the problem is probably relevant to a discussion in PNAS of the concave/convex Arrhenius plots:

Proc Natl Acad Sci U S A. 2001 Jan 30; 98(3): 848–851, "Convex Arrhenius plots and their interpretation". The theory of diffusion is evoked in this paper.

The Eyring-Polanyi equation assumes constant enthalpy of activation, entropy of activation and transmission coefficient, but heat capacity of activation induces a change in enthalpy and entropy, and eventually a curvature of plots lnk vs. 1/T.

I saw a lot of papers on the role of self-diffusion on Arrhenius plots, but in the context of solids.

Interesting. It seems to me that two activation energies are not convincing in this plot. If (!) there would be really two linear regions forming a kink, then, it would be possible to consider two temperature intervals/mechanisms with two activation energies. In opposite, the explanation by Jean-François Gal is convincing. In fact, any such model could be amended with temperature dependent activation energy, including heat capacity, and then, since a non-linearity is obvious, it is of interest to see whether the model with temperature=dependent activation energy will fit the data. In fact, such a fitting could allow obtain the change in heat capacity in reaction/diffusion rate.

It is usually difficult to determine a crossover without data over a broad range of temperature. Such behaviour is known, e.g. for planar-shaped molecular liquids where the viscosity mechanism appears to change at some specific temperature (A. J. Barlow, J. Lamb and A. J. Matheson, Proc. R. Soc. London, Ser. A, 1966, 292, 322; H. J. V. Tyrrell and P. J. Watkiss, J. Chem. Soc., Faraday Trans. 1, 1979, 75, 1417; R. Casalini and S. Bair, J. Chem. Phys., 2008, 128, 084511; H. J. V. Tyrrell and P. J. Watkiss, J. Chem. Soc., Faraday Trans. 1, 1984, 80, 1279.) For ionic liquids, the viscosity is best fit by VFT or LItovitz equations as they are "fragile" liquids in Angell's terminology. Liquids where the Arrhenius eq is satisfactory are called "strong" in contrast.The majority of hydrocarbons and high-temperature molten salts fall into this catagory. I would only consider a crossover if both the VFT and Litovitz equations showed a definite discontinuity.

Thanks for your contributions.

I will reflect on your contributions. I observe for rotational movement (estimated using NMR relaxation) at 35 ° C, according to the Arrhenius plot above, a clear discontinuity. (data unpublished)

Some examples: Phys. Chem. Chem. Phys., 2014,16, 10480-10484 and Magn Reson Chem. 2018;56:127–139.

The close contacts determined by the Nuclear Overhauser Effect also show two distinct (preferred) molecular arrangements, with 35°C the "transition" temperature between the two arrangements. (data unpublished)

According to Watkiss and Tyrrell the discontinuity may be due to loss of rotational degrees of freedom, that is a change in translational-rotational coupling.

so, probably it is a kind of phase transition, without quotation marks...second order phase transition..there might be a change in heat capacity, due to the loss of degrees of freedon, that may be determined by DSC