https://github.com/soywod/rfc2047-decoder/blob/2209238eadb0d22a543b477d09d62b880eb93c8f/src/lexer.rs#L312-L325

An encoded word with more then 75 chars should be parsed as a normal cleartext

Is it written somewhere in the RFC or is it a rule you established yourself? It does not seem right to me, if a component is identified as an encoded word (between =?…?=) then it should be treated as an encoded word, not a clear text. For example, if the length of the encoded word is greater than 76 I expect the lib to throw an error instead of returning the encoded string.