Therocephalians evolved to smaller size? Large Carnivora did not?

Brocklehurst 2019 reports,
“If these results are reliable, they support the traditional paradigm that therocephalians originated as large predators, and only later evolved small body sizes. The patterns observed in mammals do not appear to apply to therocephalians. Mammalian carnivores, once they have reached large size and a specialized bauplan, are apparently unable to leave this adaptive peak. Therocephalians, on the other hand, retreated from the hypercarnivore niche and evolved small sizes later in the Permian.”

Brocklehurst’s cladogram
posits that Therocephalia and Cynodontia arose as sisters from a last common ancestor: Biarmosuchus. In the therapsid skull tree (TST, 67 taxa, Fig. 4), Therocephalia (including Cynodontia) arises from Gorgonopsia (Fig. 2).

The question arises,
what is a ‘large size’ member of the Carnivora? Certainly big cats and walruses (Fig. 3) fall into this definition and do not give rise to smaller ancestors, as Brocklehurst notes. However, if the basalmost member of the Carnivora, Vulpavus, is considered ‘large’ then it breaks the ‘rule’ because it has smaller descendants in the LRT: Mustela and Procyon (Fig. 3). Talpa, the mole, is the smallest member of the Carnivora in the LRT. Talpa has been traditionally omitted from Carnivora studies while being wrongly lumped with the unrelated shrew, Scutisorex, instead.

I wish Brocklehurst 2019 had added
a few sample reconstructions to scale to help readers visualize the size ranges that he found in his cladogram. After all, the subject was ‘size’. I was unfamiliar with the vast majority of therocephalian taxa in his cladogram (Fig. 1).

Brocklehurst is correct:
once carnivores achieved large size (Fig. 3), no examples of phylogenetic miniaturization subsequently appear. Brocklehurst contrasted this with therocephalians, presuming that Lycosuchus (Fig. 2) was a basal therocephalian, rather than a basal cynodont by definition.

Remember:
Hopson and Kitching 2001 defined  Cynodontia as the most inclusive group containing Mammalia, but excluding Bauria. In the TST (Fig. 4) Abdalodon and Lycosuchus nest on the cynodont side of Bauria.

In the TST
(Fig. 4), cynodonts show no strong size trends until mammals, like Megazostrodon (Fig. 2), evolved tiny sizes. Therocephalians likewise show no strong size trends either (but then, I have not measured every taxon in the Brocklehurt cladogram, Fig. 1). Those that also appear in the TST are in white boxes, and they appear in several clades within Therocephalia.

References
Brocklehurst N 2019. Morphological evolution in therocephalians breaks the hyper carnivore ratchet. Proceedings of the Royal Society B 286: 20190590. http://dx.doi.org/10.1098/rspb.2019.0590