Coherent diffraction imaging of complex molecules such as proteins requires a large number (e.g., ~10^13/pulse) of hard X-ray photons within a time scale of ~10 fs or less. This corresponds to a peak power of ~1 TW, much larger than that currently generated by LCLS or other proposed X-ray free electron lasers (FELs). We study the feasibility of producing such pulses using a LCLS-like, low charge electron beam, as will be possible in the LCLS-II upgrade project, employing a configuration beginning with a SASE amplifier, followed by a "self-seeding" crystal monochromator, and finishing with a long tapered undulator. Our results suggest that TW-level output power at 8.3 keV is possible from a total undulator system length around 200 m. In addition power levels larger than 100 GW are generated at the third harmonic. We present a tapering strategy that extends the original "resonant particle" formalism by optimizing the transport lattice to maximize optical guiding and enhance net energy extraction. We discuss the transverse and longitudinal coherence properties of the output radiation pulse and the expected output pulse energy sensitivity, both to taper errors and to power fluctuations on the monochromatized SASE seed. Show Partial Abstract