Diagram of the network looks:

(b) When both x and y are large, i.e., both A and B are firing, w is increasing without bound. When one or both of x and y are small, w does not change or changes slowly. This may be seen as formalization of a stable learning.

(c)  If A and B keep firing together in time during a lifetime, w will increase without bound. This behavior cannot be biologically sustained because of synaptic normalization mechanism:

there are constraints that limit the magnitude of the weight to some value, there is a competition between the input neurons.