One of the canonical examples of "floating non-tonal feature" is [constricted pharynx] in the article "Emphasis harmony in a Modern Aramaic dialect" (Language 1984: 1-26). There are various analyses of non-tonal phenomena which make the assumption that those feature are "floating", but are equally amenable to a segmental treatment where a specific segment has the feature, and local assimilations apply (an example being Guarani nasalization).
There are analyses if vowel harmony in Maltese by Hume and a Palestinian Arabic dialect by Kenstowicz where features ([round], I believe) are set afloat and then reassociate, though I would not presently claim that no alternative is possible without reviewing the article. Another fairly classical case for floating features is rounding in Palatalization in verb inflection in Gurage lagugages.
These are all cases where paradigmatic mobility supports the case for floating no-final features, which are maximally like the strong case for autosegmental tone.