Angular momentum transport is the principal driver of galaxy evolution by spiral patterns. Not only does it change the radial distribution of mass within the disk, especially near the outer edge (e.g. Roskar et al. 2008), but it also drives increasing velocity dispersions and radial mixing within the disk. In all these ways, recurrent transient spiral patterns have a greater impact on the evolution of galaxy disks than do long-lived waves.
It might seem that if the time-averaged amplitude and pitch angle of transient spirals did not differ much from those of a steady long-lived pattern, the gravity torques and angular momentum changes would be similar. However, the gravitational stresses caused by long-lived patterns that invoke feed-back via the long-wave branch of the dispersion relation (see ch. 6 of Binney & Tremaine 2008, hereafter BTII) transport less angular momentum because the complicating advective transport term (aka "lorry transport" Lynden-Bell & Kalnajs 1972; BTII, Appendix J) greatly reduces the angular momentum flux. Furthermore, since evolutionary angular momentum exchanges between the stars and the wave take place at resonances, large changes induced by a long-lived pattern would be confined to a few narrow resonances, the most important of which, the inner Lindblad resonance, is "shielded." On the other hand, multiple, short-lived disturbances having a range of pattern speeds over the same time interval share the exchanges over large parts of the disk because their resonances are broader and have more numerous locations.
Sellwood & Carlberg (1984) showed that recurrent transient spiral patterns fade after about 10 galaxy rotation periods because the time-dependent gravitational potential fluctuations caused by the spirals themselves scatter the disk particles away from circular orbits at broad resonances (Carlberg & Sellwood 1985). It becomes harder to organize coherent spiral density variations as the velocity dispersion of the disk particles rises; spiral activity in a disk of particles is therefore self-limiting. They also showed that spirals could recur indefinitely only if some of the effects of gas dissipation and star formation were included in the calculations (see also Carlberg & Freedman 1985 and Toomre 1990), and the results from modern work are consistent with this picture. The build up of random motion is resisted in the dissipative component, and the formation of new stars with small peculiar velocities at the rate of a few per year is enough to maintain spiral activity in the entire disk.
Spiral patterns are prominent only in galaxies containing significant gas from which stars are forming; lenticular galaxies are mostly gas free and have little in the way of spiral features. Recurrent transient spiral behavior offers a natural, though not the only possible, explanation of this fact.
2.2. Age-velocity dispersion relation
The velocity dispersions of solar neighborhood stars reveal an increasing trend when they are arranged by age (Wielen 1977; Nordström et al. 2004) or, for main-sequence stars, by color (Dehnen & Binney 1998). I am unqualified to contribute to the on-going discussion (e.g. Holmberg et al. 2007 and references therein) of the accuracy of stellar age estimates adopted by Edvaardson et al. (1993) and Nordström et al. (2004), but errors of order unity would be required to totally vitiate the dynamical significance of their results. Possible smaller errors of up to 20% in individual stellar ages do not affect the dynamical implications of their results.
Spitzer & Schwarzschild (1953) proposed scattering by massive clouds of gas, before even their discovery, as the dynamical origin of the general increase of random motion with age - the age-velocity dispersion relation. Lacey (1984) extended their analysis to 3-D and concluded (see also Lacey 1991) that the observed molecular clouds could not account for the magnitude of the increase to that of the oldest stars.
Stars are also scattered by transient spiral waves (Barbanis & Woltjer 1967; Carlberg & Sellwood 1985). Lacey (1991) concluded that transient spirals were a promising mechanism to account for the higher dispersion of the older stars, although other ideas may not be ruled out.
2.3. Dispersion of stellar metallicities with age
Edvaardson et al. (1993), and others, have reported that older stars in the solar neighborhood have a spread of metallicities that is inconsistent with the idea that they all formed at the solar radius from gas that was gradually becoming more metal rich over time. This result is more critically dependent on age estimates, although errors would have to be of order unity to reduce the reported spread to a tight correlation, which would also make the Sun a truly exceptional star.
Sellwood & Binney (2002) showed that radial migration of stars driven by recurrent, transient spiral waves in fact provides a natural explanation for the metallicity spread of stars with age. They showed that the galactocentric radius of a star can migrate by up to 2 kpc in either direction as a result of "surfing" near the corotation radius of an individual spiral pattern. The combined effects of multiple spiral patterns leads to radial mixing of stars without an associated increase in the velocity dispersion. Long-lived spiral waves would not achieve a quasi-steady diffusion, since stars on horse-shoe type orbits near corotation would alternate between two mean radii, preventing the radial diffusion that is needed to account for the increasing range of metallicities with age.
Precisely the same mechanism that causes radial mixing of the stars creates large-scale turbulence in the ISM. Gas in the vicinity of corotation is driven by a spiral pattern, radially inwards by ~ 1 kpc at some azimuths and outwards by a similar amount at others. The radially-shifted gas eventually mixes with other gas (e.g. Sellwood & Preto 2002, fig. 9) at its new radius.
Spiral-driven mixing in the ISM may also help with the well-known
problem posed by the large-scale (ordered) component of B-fields
in galaxies (e.g.
Rees 1994).
Standard 
-dynamo theory
(Parker 1955)
is thought to yield too low a growth rate to achieve the
present-day observed field strengths
(Beck et al.
1996)
from the likely seed fields. The growth-rate is proportional to the
geometric mean of the rates of galactic shear (the
term) and cyclonic
circulation (the term)
(Kulsrud 1999).
Current estimates of
the -term are based on
supernovae-driven turbulence
(Ferrière 1998;
Balsara & Kim 2005),
but spiral-driven turbulence
should enhance the -effect
substantially, and thereby increase
the growth-rate obtainable from the dynamo.