The testing was at Wright-Malta Corp and was run by a PE.
He wrote a report with recommendations that has been revised a couple times
There were many thousands of connections under long-term testing.
In addition to expansion, a major problem with aluminum wire is surface oxidation. A thin oxide layer rapidly forms, and it is an insulator. In a wire nut connection there may be little actual contact between the wires because of the oxide. Most of the contact may wind up being the metal spring in the wire nut, with a couple turns carrying the current. The spring is not intended to carry the current, and at relatively high current through the connection the spring can get red hot (probably what is happening in fig 1 in the report).
The glowing spring will not trip an AFCI because there is no arc.
The author recommends, in general, applying antioxide paste, abrading the surface to remove the oxide, and making the connection with enough antioxide paste to protect the wire.
(The last instructions I saw for lugs on large aluminum wire were to wirebrush the wire and apply antioxide paste. A utility lineman said they were supposed to wire brush aluminum wire used in mid-span splices in their wire.)
The report has a couple pages on Ideal 65 wire nuts. They have an antioxide paste, but that does not fix problems with oxide already on the wire. Based on information provided I would not use them.
Alumiconn connectors were not in the original testing (they did not exist then) and initial results have been added to the report. They look like the best way to splice Al-Al or Al-Cu. Alumiconn uses set screws, and likely deforms the wire which can break through an oxide layer. Deforming the wire is likely why connections on large aluminum wire do not have the same problems as 15 and 20A branch circuits.
The new aluminum wire is harder, and not likely to "extrude" in heat-cycling at a connection.
Details of testing are not included. There are some details by way of what testing was not done on the Ideal 65 wirenuts.
One of the complaints that came out of the testing was that the UL tests used the revised aluminum wire for tests of other components. Most of the wire actually in use is the original wire. The CPSC asked UL to revise their tests and UL didn't.
The CPSC appears to have been headed for a recall of aluminum wire, which would have been enormously expensive. In the inevitable law suit the court ruled that aluminum wiring was not a consumer product and not under the purview of the CPSC. (This may have been part of the reason the CPSC dropped an investigation of FPE breakers. They had some initial testing done which was not reassuring.)