Disclaimer - I'm neither in favor of, nor endorsing plagiarism with the following question. I'm only hitting on the fact that sometimes, the interpretation of self-plagiarism can be quite stupid.
Given the well-deserved emphasis on controlling the menace of plagiarism these days, various journals and even academic institutions have insisted on various cross-check measures. Mostly, this amounts to necessitating a clean chit from some anti-plagiarism software, which (I imagine) works by comparing string lengths of some x-words in the article, with its existing database. So, if some dumb guy didn't rephrase himself, there would be common sentences, which earns him disrepute and he gets tagged as a self-plagiarist.
Now, as a ''responsible'' author, I would try to minimize such overlaps, ideally to zero. But sometimes, you can't help it. Take this context for example. (Sidenote - I'm basically a Physics.SE user.) We have used one model in two different contexts, so there is no question of overlapping content between two articles. But, since it is the same model, when I describe it, in one place I write -
The free parameters of the wawa model, p1, p2 and p3 are fitted to baryon masses and vacuum characteristics in the wawa limit.
(''wawa'' = whatever)
Now, I don't see any self-plagiarism in repeating this one sentence in the second article, but if I don't, maybe I'm being the ''dumb guy'' in the previous paragraph. So, I'll try to work around this, finding synonyms, trying alternative descriptions, but even with all my maneuvering, that sentence was the best way to describe it.
My opinion is that, even though I'm not plagiarizing anywhere in the above context, this process is turning out to be a nuisance for me. Am I supposed to sit down and waste so much time rephrasing my sentences, when I have some meaningful information to communicate to the scientific world? (Worst still, I could've been investigating some hot problem in my discipline, where urgently communicating is invaluable.)
Also, as @mhwombat hit on in a comment, am I not compromising on the best way of putting it across, when I deliberately rephrase it, just because there is a ''plagiarism'' checker in place? That's surely not what the purpose of plagiarism check was?