There is no significant tendency for the rhotic of languages to delete or "reduce" (I suppose to a glide), although it does happens in some languages. Compare this to some other common consonantal tendencies, such as spirantization of g and similar spirantization / loss of p, deletion of laryngeal glides (or glides in general), neutralization of θ, ð with other sounds (stops or sibilants).
You refer to the variable phonetic realization of the rhotic in Germanic, which can be [ɹ, ɾ, ʀ, ʁ, ɽ] and allegedly [r] (I have never heard a native speaker of English actually use [r] in ordinary speech, but you can encounter it in old movies). This is somewhat replicated in Romance where the rhotic may be phonetically [r, ɾ, ʁ, χ]. It seems that the phonetic realization of the rhotic is flexible from a diachronic perspective, but this does not easily result in a major change in phonological analysis. See this paper by Chabot, which surveys the phonetic instability and phonological stability of rhotics.
Claims about rhotics that are based only on the fact that an author used the letter ([r] as opposed to [ɹ, ɾ]) should be taken with a grain of salt, because the letter r is very often used for any rhotic. Where there is a contrast, or a detailed and trustworthy phonetic description of a language then we can conclude, for instance, that Eastern Norwegian has [ɾ, ɽ]. Claims about crosslinguistic patterns of a specific rhotic (the non-retroflex voiced lingual trill represented by IPA [r]) therefore should be scrutinized carefully, because many cases of the letter r are not IPA [r].