This seems to be the implementation: https://github.com/wuxb45/LMDB-SLB

Looks like it is effective for small data sizes and pure read-oriented workloads with heavy key distribution. Not sure whether this could work well for writes and uniform random distribution.