Actually, the U3S is a lot more than a VFO. Much of its functionality would be wasted in such an application, and it has ergonomic limitations.For a transmitter VFO, either kit could, in principle, do the job. However, there is one operational difference worth noting: the N3ZI kit is tunable with a knob, whereas the U3S must be programmed with pushbutton switches...not a very "variable" VFO. The basic QRP Labs Si5351A synthesizer board under control of a microprocessor might be a better option if only a VFO is desired; and in fact, they do offer a VFO/Signal Generator Kit that is exactly that, complete with rotary encoder and knob.
Receiving is a tougher job. In terms of spectral purity, as we've discussed before, receiver sensitivity and dynamic range are highly dependent on L.O. noise. The discussion by G3ZIL in one of the articles at the QRP Labs site shows that the Si5351A chip (with a really, really clean power source) is only about 6 dB noisier than a reasonable quality crystal oscillator. In practice, that means a carefully designed receiver made with that chip as the L.O. might be able to achieve sensitivity in the -120 dBm range--not the more desirable -130 dBm or better, but not terrible.
On the other hand, the N3ZI kit uses an AD9834 DDS chip, whose noise performance is limited by its 10-bit internal DAC to a not-at-all-impressive 60 dB broadband. Noise within ±200 kHz is better by several dB, but only in the most extraordinary cases with select devices does it approach the -74 dBc value optimistically given in the N3ZI web page. Plenty good for a transmitter...pretty lame for a DXing receiver.
John