Today, I have found a new CDC algorithm, FastCDC. It was presented at USENIX 2016 conference (a week ago) and it is considered to be a much faster replacement for Rabin-based CDC, which is used in Librevault.
FastCDC is about 10× faster than the best of open-source Rabin-based CDC, and about 3× faster than the state-of-the-art Gear- and AE-based CDC, while achieving nearly the same deduplication ratio as the classic Rabin-based approach.
Librevault Metadata Format has an extension point here especially for making it possible to replace chunking algorithm transparently while maintaining backwards compatibility with other algorithms.
Maybe, I should consider implementing FastCDC along with Rabin CDC?