tl;dr: I think there could be room to do this. However, I don't think a conclusive answer can be had through analyses of magnitudes on envelope-backs. A real answer would only come from even more detailed Monte Carlo calculations than those already outlined in Stationkeeping Monte Carlo Simulation for the James Webb Space Telescope. Sounds like a fun project!
Let's look at this systematically using well-sourced facts.
Thrust from photon pressure on sunshield
A photon's momentum $p$ is just its energy divided by the speed of light $E/c=h\nu/c$, so the force resulting in the perfect absorption of photons would be
$$F=\frac{dp}{dt}=\frac{1}{c}\frac{dE}{dt} = \frac{P}{c}$$
where $P$ is the total power of the light hitting the absorber, in units of Watts for example, and $A$ is the area of the incoming light field intercepting the absorber.
Since the sail is reflective rather than absorbing, there's a second beam of reflected light and a second force, and this one has a direction based on the orientation of the mirror. Let's just look at the magnitudes for now though.
Wikipedia gives the shape of the diamond-shaped sunshield as about 21 by 14 meters (the diagonals). That will have an area equal to half the product of the diagonals, or 147 m^2, agreeing nicely with Stationkeeping Monte Carlo Simulation for the James Webb Space Telescope.
As shown in Figure 6, the effective area of the Sunshield in the Sunward direction can vary between 105 and 163 m², the range of allowed spacecraft attitudes that prevent the telescope from being exposed to stray light.
The solar constant is about 1360 W/m^2 at 1 AU, but the L2 area is 1% farther, so let's use
$$P_{max}=A \times \text{1330 W/m^2} \approx 196 \text{kW}$$
to get
$$F_{max} = 2 \frac{P_{max}}{c} \approx 1.3 \text{mN}.$$
Acceleration is force/mass. Using 6500 kg from Wikipedia:
$$a_{max} = \frac{F_{max}}{m} \approx 2.7 \times 10^{-7} \text{m/s²}.$$
A year has about 31.6 million seconds, so that's 6.3 m/s per year of delta-v available in the +z direction if the shade points mostly back towards the Sun, and somewhat less if a bit of tilt is used if perpendicular acceleration is needed.
JWST's known station-keeping budget
Stationkeeping Monte Carlo Simulation for the James Webb Space Telescope tells us:
The results of the analysis show that the SK delta-V budget for a 10.5 year mission is 25.5 m/sec, or 2.43 m/sec per year. This SK budget is higher than the typical LPO SK budget of about 1 m/sec per year, but JWST presents challenges that other LPO missions do not face. The Endof-Box analysis was critical to the JWST mission, because it provided a realistic value for the SK delta-V budget when it was needed to establish a complete spacecraft mass budget.
So the sail provides more than double the magnitude of the station-keeping delta-v.
SOHO is an example of a spacecraft in a halo orbit (around L1) and per Roberts 2002 (from Is this what station keeping maneuvers look like, or just glitches in data? (SOHO via Horizons)) it uses a station-keeping strategy of only thrusting in the z direction (toward or away from the Sun). However, Stationkeeping Monte Carlo Simulation for the James Webb Space Telescope tells us:
In LPO dynamics is known that the x-y plane contains the stable and unstable directions, while the z direction is neutrally stable. Because JWST does not need to remain near a reference orbit, during SK maneuvers there is no need to thrust in the z direction, and the thrust vector is chosen to lie in the x-y plane.
This doesn't mean that in our non-telescope-mode, survival holding pattern we would also need the station-keeping (SK) thrust vector in the perpendicular x-y plane though, and I propose that in survival mode one could use some combination of modulation of the z-component and adding the x-y component by tilting and angling the sunshield within its safe limits will provide enough delta-v and flexibility in its direction to perform the station-keeping.
Conclusions
- JWST will experience a steady delta-v of about 6 m/s per year due to the constant photon pressure of sunlight reflecting back from its sunshield.
- While this is of course already figured into it's orbit, this will mostly result in a halo orbit just slightly in front of (sunward-of) the halo orbit about L1 calculated without the effects of photon pressure. Here "slightly in front of" is probably of the order of a few kilometers or tens of kilometers only.
- Aggressive tilting of the sunshield within safe limits can both modulate the +z acceleration, and add a component in the x-y plane
- Rotating the spacecraft about the +z axis of the orbit in the rotating frame with a tilted sunshield will direct the component of the thrust within the x-y plane, though probably not enough to make up the full 2.4 m/s per year currently obtained from propulsive maneuvers every 21 days.
Momentum unloading
I haven't through too much about how to do momentum unloading of JWST's momentum wheels using only solar photon pressure. The wheels will be needed to not only maintain attitude but also to execute regular tilts and rotations needed to direct the photon pressure for station-keeping.
As soon as the spacecraft tilts a bit, the center of the resulting photon pressure will not include the spacecraft's center of mass, so there will be at least some torque to work with.
It is possible that these attitude maneuvers can be designed in pairs to be angular momentum-neutral such that they naturally cancel each other in terms of rotations of the wheels over time.
Opinion
I think there could be room to do this. However, I don't think a conclusive answer can be had through analyses of magnitudes on envelope-backs. A real answer would only come from even more detailed Monte Carlo calculations than those already outlined in Stationkeeping Monte Carlo Simulation for the James Webb Space Telescope. Sounds like a fun project!