circuit #2 is from around 1963. the circuit is drawn correctly with opposed collectors. the whole point of the circuit is the voltage gain of a common emitter amplifier combined with the simplicity of symmetrical push-pull outputs. the original devices in the schematic were actually germanium transistors. the design actually works quite well, even though the idle current in my final modification is still a bit high. unlike the circuit in the earlier post (the class D amp, which turns out to be from about 1958 or so), the amplifier actually has a wide tolerance in output device characteristics and component values. the class D amp in post #2 is on the edge of not working, and changing component values more than a few percent stops it from working at all. the circuit in posts #5 and 6 can actually operate directly to a speaker, but i left the transformer there because it was in the original circuit. since the resistors from the bases to the supply rails set the idle current in the output devices, it's obvious that the 1.5V supply rails are about all these transistors can tolerate.
there have been more recent amp designs using the same opposed collector topology, usually as part of Sziklai pairs (kind of a modified darlington, but instead of using two NPN or two PNP transistors, Sziklai pairs are an NPN and a PNP. or a PNP and an NPN). whoever the author of this opposed collector amp was, it's clear they were thinking "out of the box" because as you mentioned, most transistor amplifiers were mimicking tube amps, with use of interstage and output transformers, and very similar topology to vacuum tube amplifiers.
in the mid 60s an engineer named Lin created the topology that has been in use ever since, with a diff amp, voltage amp and bipolar output stage (although the original Lin design was a quasi-complementary output using a darlington pair for the positive side, and a Sziklai pair for the negative side). various hybrids of Lin's design existed, some used a single ended input stage (like in the Dynaco amps), but eventually the advantages of the diff amp won out, as well as the completely symmetrical bipolar output stages (as soon as PNP power transistors were inexpensive enough, and could be fabricated with characteristics closely matching their NPN complement).
in modern amplifiers using the LIN topology, the amplifier is primarily an "op amp done large" because the amp has an inverting and a noninverting input, voltage amplifier stage and complementary output pair, in almost the same configuration as in an op amp.
btw, in case anybody is interested at looking at a modern version of a quasi-complementary amplifier, Yamaha's RS-202 receiver uses quasi-complementary outputs (drivers are NPN-PNP, and outputs are NPN-NPN). i guess Yamaha found that there's a market for the "vintage sound" of a quasi-comp amplifier in the receiver market.