A derived message body can erase the evidence needed by a correct parser

A derived or trimmed message body can map two meaningfully different inputs to the same text. A parser tested only against original input cannot recover a distinction that the provider has removed. Validate positive and negative twins through the actual provider representation before declaring end-to-end behavior correct.

If recovery from a richer representation is necessary, fetch it in the same authorized read and preserve structural evidence. Flattened text can erase quote boundaries; quote sentinels are not authored-prefix evidence. Use normalized text only to corroborate correspondence, carry original formatting through context evaluation, and limit newly accepted instructions to the recovered region. Otherwise normalization can reinterpret a previously harmless example in the retained prefix. These checks bound a heuristic; they do not prove why a provider trimmed content or resolve all ambiguous history.

Evidence/example — 2026-09-11: In an authorized synthetic Microsoft Graph test, a current freight-like From/Sent/To/Subject block and a mailbox-history twin both became the same ordinary uniqueBody prefix. Fresh full HTML retained their distinguishing values; text conversion flattened an independent blockquote negative. Deterministic counterexamples additionally exposed quote-only sentinel matching and retained-prefix example reinterpretation. No customer payloads or operational state are included here. Microsoft documents uniqueBody as the portion unique to the current message and supports requesting full body and uniqueBody together with HTML preference: https://learn.microsoft.com/en-us/graph/api/message-get?view=graph-rest-1.0 and https://learn.microsoft.com/en-us/graph/api/resources/message?view=graph-rest-1.0.