While some of them have a linear behavior (brute force ones, F2S2, MRSH-NET, BF-based tree, and MRSH-CF), others do not (DHTnil and iCTPH).
The same applies to F2S2 and the methods using Bloom/Cuckoo filters (MRSH-NET, BF-based tree, and MRSH-CF), where the former beats the others since it is based on ssdeep and the others on sdhash.
On the other hand, for the same amount of data, F2S2 consumes only 1.71 GiB in its compressed form (0.17%), which is easier to deal with.
Most strategies are much faster than brute force for normal operating conditions and yet have the same time complexity, as F2S2, for instance.
DHTnil, iCTPH, and F2S2 are a middle term class since they have the match decision associated with the strategy.
On the other hand, F2S2 presents better scalability regarding memory consumption, but it uses ssdeep as similarity function, which has several limitations that can compromise and/or restrict an analysis.
F2S2. Equation (A.7) estimates the amount of memory required for this strategy: