In fairness, it wasn't an altogether implausible or unlikely mechanism to assume of a local anaesthetic, that is would be a sodium channel blocker. Most are. Come to think of it, what if any are not? The other mechanism I can think of are is affecting TRPV receptors of the kind sensitive to capsicin, or to menthol. I know at least some NMDA antagonist drugs produce a similar numbing, but is it entirely centrally mediated via NMDARs, or do the likes of ket, MXE, PCP and its relatives also block Na+ channels?
Adder-with regards to naltrexone and opioid receptors, endorphins etc:
The mechanism of action of ULD naltrexone is mediated by blocking a high-affinity subpopulation of Mu-ORs, that are excitatory in nature. Weather Mu1ORs behave as typical, rewarding, pleasant opioid type drugs is dependent upon the state of MOR
G protein coupling. Typical bog standard MOR1 (technical term there

I mean the Mu opioid receptor types we all know...and usually love

, that are when stimulated by agonist ligands, inhibitory in their signal transduction, rewarding, relaxing, sedating you get the idea..) are coupled to Gi/o and form the majority of, but the lower binding affinity population of MOR1 and their Mu1/DOR heterodimers. Agonists (most of them, the common opioidergic drugs of medicine and of recreational use and ab-use) result in a delta-OR mediated recruitment of beta-arrestin2 and desensitization of the receptors (upregulation of DOR sensitivity within the heterodimers? not sure on that but it seems as though it would make sense to me), and with continued use of the agonist another change occurs, a switch in G protein coupling from inhibitory Gi/o binding to a higher affinity state in which MORs are coupled to excitatory, Gs and also excitatory effects mediated via Gβγ.
The excitatory population being higher affinity makes it possible to selectively block these altered receptor types using naltrexone in doses insufficient to induce blockade of inhibitory Gi and Go coupled MORs, and it also appears to directly increase
the density of these inhibitory populations. Naltrexone (and naloxone) was relatively recently also found to bind to filamin-a, which results in prevention of the switch in effector coupling from Gi/o to Gs, as supported by this paper:
http://paintrials.com/publications/pone208.pdf
Weather or not it also does so via this particular route to Gβγ I'm not sure. Still reading the paper.
I'd bet that a ligand acting as a dual MOR1 agonist/delta antagonist, inverse agonist or possibly partial agonist, not sure which subtype of delta-ORs is responsible for the beta-arrestin2 recruitment when in the guise of
MOR/delta heterodimeric receptors, and coadministration of ULD-naltrexone would make a mighty fine way to reverse and prevent the further development of tolerance to the MOR agonist.