It cannot be applied directly to active elements with negative resistance like tunnel diode oscillators.
In the real world, it is impossible to balance on the imaginary axis; small errors will cause the poles to be either slightly to the right or left, resulting in infinite growth or decreasing to zero, respectively.
Thus, in practice a steady-state oscillator is a non-linear circuit; the poles are manipulated to be slightly to the right, and a nonlinearity is introduced that reduces the loop gain when the output is high.
[6] Barkhausen's original "formula for self-excitation", intended for determining the oscillation frequencies of the feedback loop, involved an equality sign: |βA| = 1.
At the time conditionally-stable nonlinear systems were poorly understood; it was widely believed that this gave the boundary between stability (|βA| < 1) and instability (|βA| ≥ 1), and this erroneous version found its way into the literature.