A thesaurus is the most commonly used controlled vocabulary in digital information spaces, a highly structured, database-resident vocabulary that comprehends preferred terms, variant terms, broader and narrower terms, and related terms all at once, with every relationship reciprocal in both directions. Three variants exist, each with a different scope of ambition. A classic thesaurus supports keyword search, synonym management, hierarchical navigation, and associative linking all together, the most powerful option, and also the most labor-intensive and costly to build and maintain. An indexing thesaurus supports the indexing process itself but doesn't map users' variant search terms back to preferred ones, making it less powerful than a classic thesaurus but considerably less costly. A searching thesaurus exists specifically to improve search over free-text content, news archives, third-party content, where manual indexing would be too costly to attempt; it increases recall, sometimes at the expense of precision, and its cost doesn't scale with the volume of content the way indexing effort does.
All three thesaurus types solve a version of the same problem, but a classic thesaurus tries to solve it everywhere at once, while indexing and searching thesauri each pick one battle and fight only that one.
A classic thesaurus is the strongest fit here, even though it's the most expensive option. This scenario has two features that specifically call for its full capability set: manual indexing is already in place, which a classic thesaurus is built to support directly, and the audience spans from domain experts to graduate students still learning the vocabulary, which is exactly the situation synonym management and hierarchical navigation exist to serve, mapping a novice's imprecise search term to the precise term an expert indexer used.
An indexing thesaurus would have supported the manual indexing process itself, but without synonym management, a graduate student searching with an imprecise or informal term wouldn't get mapped to the correct preferred term, undermining exactly the population this publisher most needs to serve well. A searching thesaurus would have been well suited to a large free-text archive where manual indexing wasn't feasible, but that's not this situation: subject librarians are already indexing manually, so a searching thesaurus would be solving a problem this publisher doesn't actually have, while leaving the real cross-terminology search problem unaddressed.The limited maintenance budget is a genuine constraint worth taking seriously, but it argues for planning realistically for a classic thesaurus's ongoing cost, perhaps scoping it narrower at first and expanding over time, rather than choosing a cheaper option that wouldn't actually solve the core problem this specific audience has.