It’s as if there’s some kind of interaction/“additive” effect going on between these groups/tracks.
When you sum two channels, the resulting waveform is a sum of the two added waveforms.
three people talking simultaneously doesn't to me appear as one person talking three times as loud/yelling
Yes, this is a bit nonintuitive. If you add two exactly identical signals, the resulting signal would be louder by 6dB. However, adding two uncorrelated sources of the same volume like two different people talking, or even two instruments, even if they play in unison, results in loudness increase by 3dB. Three sources of the same volume mixed together are 5dB louder than a single source. Note here I'm speaking about averaged loudness, the peak volume may occasionally go higher, causing risk of clipping.
You may want to read about decibel scale. It's logarithmic and might be not intuitive at first.
On the contrary, what people intuitively call "twice louder" corresponds to around 10dB of difference.
Nevertheless, coming to your main question: a single source was below the clipping level, but summed with another source it went above. Possible solutions:
- roll down the volume. If you are not mastering the recording, this might be the preferred way. Make sure your track doesn't clip and let the mastering engineer set the final volume of the track
- apply dynamic compression or limiting either to individual tracks or to the sum. Compression automatically reduces the signal volume when it crosses a threshold you define. Compression may actually improve the sound quality if used correctly, it may also cause several damage if set wrongly, or applied excessively. There are plenty learning materials in the internet.