Based on looking at Nick's mobi_split.py code, it seems that the Mobi Header actually has a pointer and count to the SRCS record:
srcs_index = 224 (or 0xe0)
srcs_count = 228 (or 0xe4)
So I think we need a new version of kindlestrip.py that once it removes the SRCS section, it modifies the mobi (section 0) header to set 0xe0 to 0xffffffff and 0xe4 to 0.
We probably need to do that (or at least check) inside both the mobi7 header and the kf8 mobi header. We should also probably back-port this change to the original kindlestrip.py as well.