Flyback or resonant, yeah. Flyback itself can be square-pulse, quasi-resonant or active-clamp. Resonant (push-pull LLC?) would also be doable, and maybe something more unusual like current-fed push-pull.
The main restriction is, because Zo is ~200 ohms, it's relatively high (transformer windings tend to be in the 50-100 ohm range), so capacitance will dominate. Not too badly at all, this is a pretty modest ratio -- but it's worth consideration. Likewise, the output diodes' capacitance, which might be low 10s pF, reflects as much more from the primary side. So you tend to want a simple cap-input rectifier here, as opposed to like a forward converter's choke-input filter. And a forward converter would need higher Z still (generating say 2500 or 3000V at the secondary, to be filtered down to 2000V DC), further discouraging it.
Direct boost is a terrible idea because there are few transistors to even make 2kV from; a tapped boost winding could be used, but you might as well go for the transformer to get a little more design freedom -- and isolation if it helps.
Don't underestimate safety, obviously, but EMI either: even just touching wires to the output, without closing a circuit, emits a 2kV ESD impulse. Let alone fault conditions (shorting). Common ground, or isolation, may both be good strategies to deal with it; either way, that current and EMI is going somewhere, and it needs to avoid active circuitry.
Tim