They can but doing so reduces their operational life time so they'll want to minimize the amount of the thrust budget they expend on slot filling launches. As someone else pointed out their prices are high enough to cover the cost of a reusable launch with maybe half the ports filled so they're probably just playing the numbers game assuming there won't be a large coordinated drop out of customers. And those prices don't include the main top port which will likely cover a fair portion of the operation cost of the launch by itself.
The Starlink satellites are using hundreds of m/s raising and lowering their orbit. The sort of plane change you're talking about would take something close to a thousand, conservatively.
They're only going up to 84 degrees inclination [0]. Not surprising there's an absolutely tiny number of customers that would use it if they added enough satellites to make 100% coverage.
Hmmm, that is a relatively high delta-v to change inclination that much (~1.5-2 km/s for a ~12-15 degree difference). On the other hand, if the shared payloads are light enough (and they very well may be, particularly if the launch is undersubscribed), the upper stage could do the orbital transfer and should have plenty of delta-v to do so (particularly with a droneship landing) considering F9 regularly does recoverable flights with many tons through 2.5km/s above LEO.