If you consider Rice (Oryza sativa), then aroma is definitely a crietrion (but not alone) for varietal designation. Presence/ absence of Aroma is an important descriptor of rice landraces and farmer varieties. You may refer to INGER guidelines for rice characterization.

in essential oil rich Labiatae species smell (composition of e.oil) is often used as a chemotaxonomic marker, but rarely are taxa delimited by it. (!!!)

aroma is used as an important character in some taxa for diagnosis along with other characters (morphological, in particular), but it can't be the only character to delimit taxa. variants of the same species with different aroma must also have some morphological and other differences e.g different rice varities. Hence it can't be a reliable dignostic character.

In orchids that produce fragrance, some species show variation in the aroma that depend of factors as light intensity, temperature, humidity or other. The plants of one species can produce a specific aroma, but its constitution become different when they are subject to different environment. So, the aroma can not be useful as taxonomic character, or the only one, unless a study reveal that it is constant among all members of a population and in different environmental conditions.

Aroma is a character like any other, and like all characters it depends on the context. Consider a case where in the field there are plants with two subtly different morphotypes which co-occur and have no intermediates, and I notice that the two morphotypes have distinctly different aromas. I'd take that seriously in helping decide whether or not I had two different species

"Aroma cannot be a reliable character"... this statement is untenable in plant taxonomy. It's of course unreliable if used alone to distinguish SPECIES, but for INFRASPECIFIC taxa, aroma can be definitive and distinctive. In the case of rice, aroma alone may identify a specific varietal population - of course, if "it is constant among all members of a population and in different environmental conditions", as Rodolfo Gómez writes.

I would not consider aroma per se a reliable character, it can only suggest that a difference exist. It would be far more reliable if you could asses the chemical composition of the aroma. This would give you both qualitative and quantitative data which can be compared with confidence.

I think it is good indicator to guess major taxon as family or genus or even species but it is not authentic in its sole use, it does need some synergistic application of other morphometric characters to identify or classify species.