This is not always true. The most important factors are coherency between precipitate and matrix and size of the precipitate.
For most common example of Al-Cu system only nonequilibrium phases are coherent or semi- coherent with the matrix and this is why the maximum strengthening is obtained before equilibrium state.
In the case of nickel based alloys gamma prime is always is coherent with the matrix but the lattice misfit may vary during ageing. In general I would guess that in case of nickel based alloys low temperature properties such as hardness and strength are highest when the size of gamma prime is in range of 5-30 nm.
Maximum strengthening is obtained when precipitate is small and coherent with matrix.
Its a common phenomena in precipitation hardening, an alloy softens after attaining maximum hardness. However, the basic effects controlling the maximum hardness actually depend on a number of factors:
(a) the alloy composition
(b) the quench rate
(c) the ageing temperature
(d) the kinetics of homogeneous cluster nucleation
(e) the growth rate of clusters
(f) the coarsening reaction as ageing proceeds
(g) the precipitation sequence - the transformations that the precipitates progress through as ageing proceeds.
Thermodynamically speaking, the precipitation follows the path of minimum activation energy rather than maximum loss of free energy. In other words, the kinetics favor homogeneous nucleation with a coherent interface, followed by transformation to one or more intermediate semi-coherent phases before final transformation to the equilibrium precipitate.
To sum up, the interfacial free energy will be minimised with better matching of the two phases (deteriorate mechanical properties). Incoherent interfaces have high energy and are relatively mobile because of the greater freedom of atomic motion (maximum hardness and strength).
As several people already mentioned, the answer lies with the state of coherency between the ppts and the matrix. However, just saying this does not really reveal immediately the underlying cause. Coherency, and to a lesser degree semi-coherency, does NOT mean the matching between the matrix and the precipitate is perfect. As a result, the two neighboring phases accommodate this mismatch and what is known as "coherency strains" are created at the interface. One should remember that every dislocation also has strains fields around it (compressive stresses above the dislocation line, and tensile below). So now the dislocations interact with the stress (or strain - name it as you like) fields around the coherent/semicoherent ppts. In other words, the precipitates exert their presence over much greater distances than their actual physical sizes. In this picture, the ppts act as far more effective barriers to dislocation motion (coherent ones are more effective than semicoherent ones as the coherency strains are greater for the former)... this is something the incoherent ppts are not as capable of. Needless to say that coherency/semicoherency cannot be maintained as the ppts grow. Due to coarsening of ppts the coherency strains diminish with a concomitant reduction in hardness/strength.
As i can see all of you related the variation of mechanical properties such as strength and hardness to the coherency between precipitates and the matrix but can i relate it with thermodynamical aspects like interfacial energy ?