And remember only 1 of the 8 tones is TX at any time so the waterfall is actually a misrepresentation of how crowded the spectrum is
The biggest problem I have is when you get a strong station overlapping a weaker station then you do tend to lose the decode of the weaker station
If you run SpectrumLab (free) on a ft8 band you will see how true this is.Hi
FT8 uses all 8 frequencies for a message transmission, one at the time. The receiver looks all those 8 frequencies as it needs to detect which one is sent. Frequency spectrum of a signal is the total transmitted tones i.e. about 50 Hz for FT8. Actually it is about 3 to 6 Hz wider due to the "fast" frequency shifts depending how spectrum width is defined.
FT8 uses and needs the whole 50+ Hz bandwidth for a successful reception. Each tone as such require about 6.25 Hz bandwidth for a "perfect" transmission and reception. In that sense a second signal affects to decoding about the same amount within 6 Hz frequency difference including 0 Hz difference.
By the way there are three dimensions that affect to the decoding probability of two signals on the same frequency slot. Everybody knows signal strength and frequency difference, but also time difference has a strong effect. A time difference of more than a symbol length 160 ms may be enough even, when signal strengths are the same and there is no frequency difference.
73, Reino OH3mA