Your question prompted me to look at what had been written on the topic, specifically for plants. Well, there are all sorts of reasons for lumping genera and all sorts of reasons for splitting them. A good review article is
History and Concepts of Big Plant Genera Author(s): David G. Frodin Source: Taxon, Vol. 53, No. 3 (Aug., 2004), pp. 753-776 Available via this link: https://iks.ukzn.ac.za/sites/default/files/4135449.pdf
From a quick skim of that article I learned that the issues of “What is a genus?” and “How big is too big?” have been around since the days of Linnaeus. Linnaeus detailed 56 species of Euphorbia in the first edition of Species Plantarum, but also believed that genus should have no more than 100 species. Expansion after that point tended to proceed based on where collection was happening – mostly with genera from temperate regions at first, then with those from colonial possessions.
In general, botanists stuck to the concept that all species with a consistent “character” formed a genus. By the 1880s there were 28 plant genera that contained more than 300 species. There were various attempts to split the larger genera, but the principal obstacle was a lack of clear differences that could be used consistently. The paper looks at Rhododendron as an example of a large genus that could not easily be split and Cassia as an example where floral and seedpod form allowed separate genera of Chamaecrista, Senna and Cassia to be described.
In the 20th century, there were renewed concerns about the unwieldy nature of big genera preventing comprehensive analysis or easy keying. But (to my mind) a lot of the lumping vs. splitting arguments of this period were ultimately futile because the botanists involved had no access to data on the underlying genetic relationships. That started to changed with techniques such as karyology and since the 1980s has undergone significant shifts through phylogenetics.
Ultimately, though, it’s not just enough to say “This genus is too big”. It’s also necessary to be able to identify two or more separate clades that can be conclusively separated using a consistent combination of characters.