
Another way to track this is through the phasor diagram of the generator, using the constant-power constraint directly.
With the bus voltage $V$ fixed as the reference phasor, the real power delivered is proportional to the component of $E_f$ that is in phase with $V$, scaled by $\sin\delta$. Holding $P$ (and hence $V$, $X_s$) fixed while raising $E_f$ forces $\sin\delta$, and hence $\delta$ itself, to shrink, since a larger $E_f$ needs a smaller angle to keep the product $E_f\sin\delta$ unchanged.
Geometrically, the stator current phasor $I_S$ is found from $E_f=V+jI_SX_s$, so $I_S=\dfrac{E_f-V}{jX_s}$. As $E_f$ grows in magnitude while its projection onto the $P$-axis stays fixed (constant $P$), the phasor difference $E_f-V$ grows, so $|I_S|$ grows too.
At the starting point, unity power factor means $I_S$ is exactly in phase with $V$. As $E_f$ increases beyond this point, the operating point moves into the region where $I_S$ lags $V$, which for a generator is defined as the LAGGING power factor region, corresponding to overexcitation and the generator supplying reactive power outward to the system.
So all three changes happen together: $|I_S|$ increases, the power factor shifts from unity to lagging, and $\delta$ shrinks.
\[ \boxed{\text{Stator current increases, power factor becomes lagging, load angle decreases}} \]