This would be a major setback. And thus, describes the intent of the Trump administration - to extort a peace on their terms, as if they were the enemy - quite vividly.
For sea drones, there could be a work-around (some other satcom system) but for ordinary units on ground, a work around must be cheap - for them, engineers would be drilling posts into ground and running temporary fiber to temporary base stations like there's no tomorrow. One hop of fiber can be quite long, 80 km is no problem with really cheap COTS hardware (professional hardware can probably talk over hundreds of kilometers). It's making changes that is problematic.
Of course, units would also fall back to using civilian cell phone networks, some with directional antennas (e.g. you know that 10 km to your rear is an intact base station at (X,Y), you point a 20 decibel parabolic dish towards that direction and get online without the enemy having a good idea about your whereabouts. A big hassle, but not an insurmountable one.
Long range strike drones 99% likely aren't using Starlink. They fly by inertial navigation and machine vision, and pick up clues from Russian mobile networks.
One absolutely doesn't prepare such a large operation at such a short notice.
For an intelligence analyst, signs of an invasion are typically detectable 3 months ahead. If one performs a maritime invasion at a notice of weeks, failure is likely. (For reference, the D-Day needed years of planning and months of moving resources to work.)
Also, I trust that Taiwan has infiltrated China just as deeply as China has infiltrated Taiwan - they likely cannot keep massive secrets from each other.