He's talking about buying two matched quads. From this he will pull two matched triplets and have a spare for each triplet.
IMHO, this is a good strategy, especially if a bias probes is used to select the best three tubes out of each quad. I do agree the two quartets should be relatively close in their matching.
The only concern I have is he mentioned buying EHX6550s. Every EHX 6550 or KT88 I have ever had has failed in very short order. Sixteen tubes total in three different amps; pretty much a 100% failure rate with very light usage. Multiple failure modes: some flashed over, some developed low emissions, some stopped conducting completely, some last the ability to hold bias.