When I select output tubes for a fixed bias Fisher amp or receiver, I match pairs in the same channel. My rationale is that, in a push-pull pair, I want both sides of the excursion of the waveform on each channel to be able to go to the same or a similar amplitude before clipping.
I do my selection in circuit, with fully powered tubes and a 10 ohm resistor between cathode and ground. I check bias, swap tubes around, and then let it sit and check it again in half an hour. Some tubes, in particular the 7591, which I use a lot, can drift over time if they're getting bad. Observing that a tube drifts when under power for some time removes the tube from the queue and the next one is substituted in. I think Gordon W called it grid drift or something like that. Anyway, I am the beneficiary of his expertise on this subject.