The standard /. IANAL applies here, but I'm pretty sure that if you have legal access to the copyrighted text (ie you or someone you know owns a copy of the magazine) then it is ok to create a derivative work for the purposes of searching that work. This is the loophole that Google (name your favorite search engine here) uses, and they go so far as to offer cached versions of some sites.
Lucene, or a more friendly wrapper around it like SOLR, has the option of creating a search index based on an original text from which the original content cannot be extracted (indexed=true, stored=false on a field), so that would seem to cover the case of finding an article without violating the rights of the author or the publisher.
As for not having the text online, I'd suggest either scraping the archive sites in the process of building your search index, it's pretty hard to search something that isn't digitized.
Best of luck, as this sounds like a worthwhile project. I do think that the volume of data you're discussing would fit easily in a SOLR instance that would consume very modest amounts of server resources to operate.