As someone who is currently sitting in front of an aging laptop plugged into a pair of rather nice desktop displays, I'm a little sad about the lack of support for more than one external display with the M2. Honestly, I would probably order an M2 MBA straightaway if it were not for that limitation.
The problem isn't with the PHYs, as both the M1 and M2 have two independent Thunderbolt 3 / USB4 ports in addition to the eDP output for the built-in panel / HDMI PCON. It boils down to the little M series GPUs only supporting 2 display pipes, and the display controller not being flexible enough to route both display output streams to the Type-C ports—one is always routed to the eDP interface. On the M1, that eDP display output stream is also limited to a maximum resolution of 5120 x 4096 (or "looks like 2560 x 2048" HiDPI), whereas the one that goes to the Type-C ports can do 6144 x 4096 ("looks like 3072 x 2048" HiDPI).
It would have been sweet if the M2 had inherited the M1 Pro's 3x 7680 x 4096 ("looks like 3840 x 2048" HiDPI) output capabilities. Although I suppose Apple will sell a lot of 8 GB unified memory / 8-core GPU M2 MBA configurations which might struggle a little with three 8K displays. Other GPUs tend to combine / consume two pipes for resolutions that would require more than 25.92 Gbit/s without compression, or the equivalent of 4-lanes of HBR3. In the case of Intel's Alder Lake IGPs, that works out to a maximum of 5120 x 3200, 60 Hz, 24 bpp for a single pipe, or 7680 x 4320, 60 Hz, 30 bpp for two, and subject to memory bandwidth availability. It seems like Apple has opted for bigger pipes but fewer of them.
According to Hector Martin, "One display pipe is larger than the die area of two performance CPUs on the M1 Pro."