If you are composing you frame using chunks, you'd need to have cut outs where the other chunks go. Unless you are composing and downsampling the image 100% on the CPU, in which case you wouldn't need alpha, but you are doing work on the CPU.
If you use FSAA, your backbuffer is going to be too large for the 2ish MB, so you have to tile. If you have to tile.
You couldn't get the Gamecube's GPU to downsample a tile anyway, because it can't texture from the framebuffer. At least not without copying the data around, but to even get it in texture memory you'd have to tile the tile. And I don't even know if the 24-bit framebuffer formats are supported texture formats.
I can't find good documentation on what the video interface's capabilities are so I don't know what its maximum resolution is (performing FSAA to go below max resolution would probably not be a good tradeoff) or if it has any kind of downsampling of its own. A simple integer 2:1 downsampler wouldn't be that bad CPU-wise, although the memory bandwidth hit would be large.
Suffice it to say that the system wasn't geared for tiling,
But I'm still not seeing what dest alpha has to do with this, unless you're thinking that the function to copy from the GPU framebuffer RAM to the front buffer in main RAM can perform per-pixel alpha blending and that it'll scan every source pixel and accumulate them on top of each other..