Thermal expansion/contraction is clearly an interesting issue.
Assuming a fixed point at the midpoint between LA and SF, the station at each end would need to pick up the delta length from a 175 mile segment of steel tube. Applying Larry's assumptions from above, that means that the location of the end of the steel tube will move by ~801 feet at each station. I don't understand the station infrastructure sufficiently to propose a solution, but it should be possible at reasonable cost and functionality.
I also think that, due to the low bending radii involved, the steel would have sufficient elasticity to accommodate the bending from expansion/contraction. However, the designers have assumed that they can have sharper bends at the low velocity sections near the stations, and that is where this bending is most critical. This is easy to visualize this if you picture that there is a straight section pulling in towards the station, and then a bend when it gets there. When the steel expands, basically you would have to push the straight section of steel into the bend, and the bent steel would go in to the (presumably) straight platform area. Such movements would also create lateral moments on the pylons (which are loads they don't like) and could also create issues of steel fatigue. All of this has to be examined, and could range from "cost increase" to "show stopper".
The inside curvature of the tube is also going to be slightly affected. The diameter will be ~0.1 percent smaller in cold weather than in hot. The air hockey skis are going to be finely manufactured to match the geometry of the tube. They need to verify that it will be able to cope with this variation.