strace does not yet follow the clone system call, and it follows vfork by rewriting it to fork. SUBTERFUGUE follows these calls closely (though CLONE_PTRACE is not implemented).
strace's emulation of the wait4 system call is pretty rudimentary. It stops the waiting parent, making it insensible to most signals, it doesn't discern between waiting on process groups versus waiting on all children, and it doesn't handle the WUNTRACED or __WCLONE flags nor the rusage argument. Consequently, strace has problems following job control. SUBTERFUGUE does a more faithful emulation by rewriting wait4 to pause (thus obviating the need to stop the parent), but it also does not yet handle the rusage argument.
strace can attach to running processes. SUBTERFUGUE does not yet support this.
strace, of course, works on many different Unix platforms. SUBTERFUGUE runs only on Linux and might be impossible to port to non-free Unixes.
strace is implemented entirely in C and, at least for now, is much faster than SUBTERFUGUE.
No. It's a play on "subterfuge" and "fugue".