Ceramic capacitors are really quite rugged. They behave (and fail) like a piece of wire when ripple current is applied: The current heats the capacitor via its ESR and if it gets too hot, it gets damaged.
Let's take the C5750X7S2A106K230KB from TDK as an example. This is a 10µF 100V X7S capacitor in a 2220 form factor. TDK very helpfully provides characteristic data for this capacitor.
Looking at the "temperature rise characteristics" in the characteristics sheet, you'll see that this capacitor can take about 4A at 100kHz for its temperature to rise by 20°C. This doesn't mean that that's the maximum the cap can take, however: The relationship between ripple current and temperature rise is quadratic, as in P=I²R. If you're okay with a 45°C temperature rise (or can provide cooling to the capacitor by means of large copper pours on the PCB), you can push it to 6A.
Another thing to keep in mind is the capacitor's DC bias characteristics: Even though this capacitor is advertised as "10µF", it'll drop to about 6µF at 30V already. This means that you'll need three of those in parallel to reach your required 17µF, not just two. Characteristic sheets are really the only way to figure this out, which is why I stay far away from ceramic caps that don't provide this data. I wouldn't be surprised if your 10µF 35V cap drops to something ridiculous like 2µF when you apply your 28V to it.
An Aluminum Polymer capacitor, like the 50SEK22M from Panasonic, might be a better option. They're much cheaper than ceramics for this capacitance range and can take a lot of ripple current too. Don't forget to add a small ceramic cap in parallel, on the order of 100nF, to improve high-frequency filtering. (This cap doesn't have to take much ripple current. The current is shared in proportion to the capacitances, so about 0.5% of it goes through the parallel 100nF cap. This is negligible.)
