Blockchain (BC) technology has been established in 2009 by S. Nakamoto, using the Proofof-Work (PoW) to reach consensus in public permissionless networks [1]. Since then, several consensus algorithms were proposed to provide equal (or higher) levels of security, democracy, and scalability, yet with lower levels of energy consumption. However, Nakamoto's model (a.k.a. Bitcoin) still dominates as the most trusted model in the described sittings since alternative solutions might provide lower energy consumption and higher scalability, but they would always require deviating the system towards unrecommended centralization or lower levels of security. That is, Nakamoto's model claims to tolerate (up to) < 50% of the network being controlled by a dishonest party (minority), which cannot be realized in alternative solutions without sacrificing the full decentralization property. In this paper, we investigate this tolerance claim, and we review several approaches that can be used to undermine/overpower PoW-based BCs, even with minority. We discuss those BCs taking Bitcoin as a representative application, where needed. However, the presented approaches can be applied in any PoW-based BC. Specifically, we technically discuss how a dishonest miner in minority, can take over the network using improved Brute-forcing, AI-assisted mining, Quantum Computing, Sharding, Partial Pre-imaging, Selfish mining, among other approaches. Our review serves as a needed collective technical reference (concluding more than 100 references), for practitioners and researchers, who either seek a reliable security implementation of PoW-based BC applications, or seek a comparison of PoW-based, against other BCs, in terms of adversary tolerance.