156.25 Hz sounds familiar… SiLabs chip?
Agreed, I think a better approach is to do the same as being used for Fossasat-1, which is to use the BPSK Demodulation flow-graph, which will save USB-demodulated audio (and completely fail to demodulate any BPSK, but thats OK). The BPSK demodulator automatically aims for the signal to be at a 1.5 kHz IF in the demodulated audio, so there should be no need to adjust the RX frequency from the nominal.
The CW decoder does the same, but it will likely mistakenly decode some CW data, which will cause the observation to be incorrectly flagged as valid (known issue).
I’m pretty sure the RX will be off-frequency initially, due to TLE inaccuracies and other drift. A few stations (mine included) are set up to save IQ data, so that will help.