Should Wiktionary require that all its information artifacts are sourced from reliable sources?

The wikt:Wiktionary:Criteria for inclusion (CFI) of the English Wiktionary requires definitions of English entries to be substantiated by quotations of use, thus allowing original research into definitions and sense structures. Should it instead require that definitions and other artifacts are sourced from authoritative reliable sources such as dictionaries?

Items: definitions, quotations of use, synonyms, hyponyms, etymologies, etymologicallly related terms, thesaurus, source conflict resolution vs. original definition invention, non-derived base information vs. information derived from it (esp. binary relations).

Arguments for

 * That is the only way to ensure reliability; amateurs interpreting primary sources such as examples of use cannot guarantee reliability and trustworthiness.
 * One can in fact avoid certain errors in the "reliable" sources when one allows significant editor input into source and evidence assessment. Thus, the English Wiktionary has wikt:Appendix:English dictionary-only terms.
 * There is a difference between allowing editor assesment and judgment to override sources in certain cases vs. considering what other dictionaries say to be immaterial as the English Wiktionary wikt:WT:CFI seems to do.
 * Expanding on the above, e.g. Poldauf 1965 Czech-English dictionary has "anglistics" as an English translation of Czech "anglistika" alongside "English studies", although "anglistics" is almost never used in English, especially not in lowercase "a". One can remove "anglistics" from Wiktionary if one allows that evidence of use trumps sources considered authoritative and reliable.
 * In this case, one can point out that authoritative/solid English dictionaries do not have "anglistics" and let that fact override the lone Czech source.
 * Take the entry wikt:cat; the subsensing is pretty much not found anywhere else, is absurd, and shows what happens when not-so-smart amateurs are allowed to assert their will in a wiki.
 * That seems like a subjective assessment, hard to verify without a clear statement of what is wrong with the subsensing.
 * The argument was made in the spirit of experientialist epistemology: look, see for yourself, and make your own judgment.

Arguments against

 * Even if we grant for the sake of the argument that definitions should be sourced from dictionaries, it does not follow this should apply e.g. to "Related terms", "Hyponyms", etc. Even Wikipedia does not require its navigation boxes and "See also" items be sourced from reliable sources.
 * For a language that has a dictionary supporting definitions or at least translations to another language but has no synonym dictionaries, the editors would be disallowed to enter synonymy based on definitions. It should be part of general semantic intelligence to determine whether words are synonymous based on their definitions.
 * Even for definitions, reliable or solid mainstream English dictionaries do not always have same sense structures and equivalent definitions. Ultimately, the editors have to insert their assessment of sources and evidence and certain degree of discretion to shape the senses and their definitions in Wiktionary anyway.
 * Sure enough. However, there is a difference between inventing definitions from citations of use (full original research) and resolving conflicts between solid sources (partial original research).
 * Requiring definitions to come from solid sources containing definitions or at least translations would prevent Wiktionary from having words and phrases not defined anywhere else, a loss of what is in fact unique differentiator of Wiktionaries.
 * Expanding on the above, even the very meticulous Germans allow substantiation of definitions via quotations of use in the German Wiktionary.
 * Assignment of a quotation of actual use from literature under a specific sense is also an amateur interpretation, running the risk of being wrong. Upon strict no-original-research interpretation, Wiktionary would have to defer to reliable sources for quotation selection; for some small languages, it would have to give up on providing quotations of use entirely.
 * Granted. On the other hand, there is quite a difference between the severity of error of quotation misassignment vs. error in a definition.
 * A Wiktionary for a language much smaller than English would often not be able to directly source translation pairs between the language and a small target language. By contrast, one can reasonably reliably fill translation tables if one allows synthesis in the sense that one combines multiple sourced or well established statements into a new statement that is not directly sourced. Thus, one can describe words in various languages using a lingua franca, nowadays English, obtain relevant statements about the words in the lingua franca, and then translate the result into the local language of the specific Wiktionary.
 * The English Wiktionary thesaurus would need to be abandoned since it is not directly sourcable from authoritative sources, only indirectly. This pertains especially to the "Various" section which allows out-of-the-box semantic relations, but also to "Synonyms" in so far as the synonymy is mostly derived from mainspace definitions rather than being sourced from dictionaries of synonyms. The reader is better served with this kind of semantic navigation aid that is somewhat unreliable (which wikis are anyway, try as one might) than an empty web page. In case of doubt, one could place unreliability disclaimers on the thesaurus pages.
 * One can draw a distinction between binary relations that can be derived from other reliably sourced information, and the primary information from which other things derived. Thus, synonymy and related terms would be derived while definitions and etymologies would be non-derived. And thus, the thesaurus would be considered derived, with no need of direct sourcing, while definitions would be considered non-derived, to be properly sourced.
 * Requiring translations in a translation table to be sourced would seem to burden the page with citations, eventually as many as 200.
 * On the other hand, one could require each translation pair to be authoritatively sourced from the non-translation-table target page. Thus, for (cat, Katze) in the English Wiktionary, the sourcing would not be in the "cat" entry, where the pair is present, but it would be in the Katze entry.
 * One could understand/interpret "be sourced" in a different way: not that inline references need to be in place but rather that they need to exist in principle (perhaps easy to find via Google search) or that they need to be in an appropriate page, e.g. talk page or a dedicated namespace meant for extensively detailed sources/references. In the dedicated namespace implementation, the verifying person would combine the mainspace information from a translation table with the separate source list information; while less convenient with less granular traceability than inline references in the mainspace, it would still do the job of proper sourcing.