Sci-Hub's an important player in the transition to Open Access. Note, however, that many papers are already available elsewhere (e.g. ArXiv), legally. You can use the OAButton [1] or Unpaywall [2] extensions if you hit a paywall to find a free version. They're not perfect, and solve a problem that shouldn't exist, but it's nice that they're there. [1] https://openaccessbutton.org/ [2] http://unpaywall.org/
The great thing about Sci-Hub, and a not-inconsequential element of its success, is that it is very nearly universal. Content is simply available in the archive in a tremendous number of cases. And is directly referenceable by DOI (or, when it's working, direct search).
The size of an archive matters as it reduces search costs across archives (this is a reason why archives tend to "a single max-size dump" dynamics). If you look at lists of the world's largest libraries, for example, it's pretty much the U.S. Library of Congress, and ... everything else. Even at the university library level, the largest collections tend to be fairly uniform at about 15-20 million volumes (Harvard, University of California, etc.).
The fact that a scholar can go to one such institution and have access to their entire archive is a compelling advantage. Shoe-leather adds up when you're crossing provincial or national borders.
Similar dynamics drove the adoption of single scholarly languages -- Greek, Latin, Arabic, Latin (again), French, German, and (in a battle with German and Russian) English following the 2nd world war, given that works had to be translated only to one language (English) rather than mulitiple.
Similar arguments, compounded by the insane costs of scribe and codex formation, applied in the pre-print era.