Ok, so I have finished reading O'Brien and JJ Goedert. I think the accidental lab worker cases probably have low utility in distinguishing between pathologic HIV and passenger HIV theories, but the animal SIV studies potentially fulfill the ideal experiment design. If animals injected with pure close SIV analogs to HIV develop AIDS-like illness, that is about as good as evidence as one could hope for.

For the lab worker infections, they say that the genomes are close to the clonal strain used in the lab, but they don't have a good probalistic model for this:

The sequence divergence between LW virus envelope genes and clonal HTLV-IIIB is <3%, which is the same genetic distance from LAV-LAI to HTLV-IIIB, viral strains now agreed by all to have been derived from a single patient [54,55]. This low level of sequence divergence is equivalent to the variation observed between HIV-infected infants and their mothers, and threefold less than the extent of variation of HIV between unconnected patients [56].

threefold more or less variation doesn't sound like much to me, and regardless even assuming they did all get infected via the lab, it doesn't mention treatment, so we don't know if they were treated with AZT or what.

I think the animal models are more interesting - as all the drug cofounding variables are eliminated and everything is controlled.

The authors discuss 3 animal models - baboons, some wierd knockout mice with human t-cell graphs, and finally SIV injected into rhesus monkeys. They spend the most time discussing the latter paper so I looked into it.

"Induction of AIDS in Rhesus Monkeys by Molecularly Cloned Simian Immunodeficiency".

In short, all 5 of the monkeys injected later become seropositive, and 50% of them progress to AIDs-like illnesses and die. Presumably some are naturally resistant.

There is a potential issue to this otherwise smoking gun, which is that it appears they had to inject whole culture of T-cells, as the virus can not be isolated directly into plasmid vectors:

We, and others, have experienced considerable considerable difficulty in subcloning the full-length proviral DNA into plasmid vectors.

SIV mac239-cloned DNA was transfected into primary macaque PBLs and into Hut 78 cells (a human CD4+ T cell line) by a DEAE-dexxtran procedure, and stock virus was frozen for subsequent animal inoculations

But if all they had then was provirus integrated into PBL ( peripheral blood lymphocytes ), then one must wonder about the risk of graft versus host disease, and transfusion reaction in general. Basically, the foreign T-cell line can actively invade and cause all kinds of havoc. Perhaps they have controlled for that and I'm completely off, but I don't see where they mention it.

OBrien and Goodart conclude:

Control macaques inoculated with saline, with inactivated SIVMAcA239 or with SIVMACA239 carrying mutations/deletions in the nef gene failed to induce AIDS in the same macaque species [80,83,84]. Because SIV strains cause AIDS in monkeys and because they are the closest phylogenetic relatives of HIV, they provide an animal model fulfillment of Koch's transmission

I haven't seen mention of 'inactivated' sivmaca239 yet in 80, but maybe it is in 83 or 84. However, if they are forced to infect live T-cells because they can't produce viable clonal plasmid, this seems to have a major confounding factor in graft vs host disease. I'd like to find a knowledge skeptic take on this - perhaps Duesberg confronts it, not sure yet.

If he doesn't confront it, then that doesn't look so good for his position.