There is no clear quantitative cut-off point, if that's what you're looking for.
The severity of plagiarism has to be evaluated qualitatively and within the context of the work itself and of the discipline and its conventions (e.g. regarding citations and "public", unattributable knowledge). Qualitatively here means several things:
- Has the author fraudulently claimed a major idea or a finding? Even with perfect paraphrasing, this -- it might be argued -- is the most severe form of plagiarism.
- Has the author claimed an idea as original that is only tangential to the author's contribution (but doesn't bear much argumentative weight)?
- Is the author merely sloppy with citations, i.e. can it be guessed from the context that the passage was meant as citation, that the attribution was (conveniently) forgotten; or is it merely incomplete or erroneous?
- Are we talking about "self-plagiarism" (if that exits)?
Depending on the answers to these questions, 50 examples might be a scandal or not worth making a fuzz about.
Moreover, even if plagiarism could be evaluated only quantitatively, it is unclear what you would measure. The number of copied and unattributed passages? The length of each problematic passage? The ratio of changed words in a paraphrase?