I'd like to chime in here and suggest that a fundamental problem here is that the capacitance load needs a small series ballast resistor.
An amplifier's output impedance is established by its open-loop Z-out divided by its open loop gain. We know the loop gain falls with frequency, e.g., at -6dB/octave, so the closed-loop Z-out rises with frequency. In fact, the output looks and acts like an inductor. Adding a large capacitive load creates an L-C resonance, with a peak at the resonant frequency, Fres. The O.P. needs to add a series resistor to his capacitor, of about Rs = 1 / 2 pi Cload Fres. This resistor is located after the ampliifer's feedback connection point, and Rs Cload will set an upper -3dB response rolloff frequency for driving the capacitive load. If this response-limit frequency is inadequate, other measures will be needed to reduce open-loop Z-out.
Just to add a comment: Comparatively speaking, MOSFETs have a fairly low transconductance, e.g., compared to a BJT. This creates a fairly high open-loop Z-out, that the LM49830 overcomes with its loop gain. We shouldn't blame the the LM49830 for any troubles that ensue from the MOSFET's inadequacies. :-)
Cheers,
- Win
