as you know, one call is using one specific frequency during one/eight of the time.
if somewhere else in the area, another call is using the same frequency during the same time, the first user will hear a degradation in the voice quality (robot voice).
this is co-channel interference.
2 calls using the same frequency, in the same area.
if there are 8 users on the first frequency, and 8 users on the second frequency, you see that the probability of “collision” is very high.
now, same scenario, except that the second user is using a frequency which is +200kHz or -200kHz compared to the 1st frequency. Here, the impact on the voice is much less : only if this “adjacent” frequency is very very strong, then the first frequency will be “eclipsed” and the first subscriber will hear a bad voice quality (still the robot voice).
the cell doesn’t identify the interference, it is up to the optimizer to find which frequencies are conflicting.