Same. If nothing else zstd universally supports compression and decompression with threading across any number of cores. It will even adjust threads dynamically

Same. If nothing else zstd universally supports compression and decompression with threading across any number of cores. It will even adjust threads dynamically based on the next limiting factor (I/O speed). Meanwhile it's also generally faster and provides significantly better compression ratios (as you describe). In my mind "always threading, everywhere" is reason enough to always use zstd. Sure there are implementations for other compression algorithms that support threading but the reference zstd implementation from Facebook always does, everywhere.