I think Korean doesn't traditionally have word-medial obstruent + sonorant sequences (excluding semivowels, which can be analyzed as part of diphthongs). Nasals cause a preceding obstruent to become a nasal by assimilation. The non-nasal sonorant ㄹ doesn't occur word-initially in native Korean words nor, I believe, at the start of Korean morphemes, and in Sino-Korean it seems to be replaced with /n/ when it is preceded by a consonant. See the examples in this paper: "Decomposing the Syllable Contact
Asymmetry in Korean", Karen Baertsch and Stuart Davis
In Korean [...] If the two consonants at the syllable boundary have falling sonority then the sequence is normally maintained. That is, when the consonant at the end of the first syllable is more sonorous than the consonant at the beginning of the following syllable the sequence is maintained as exemplified by [kal.pi] ‘ribs’, and [cal.mot] ‘fault’. However, if the sequence potentially has rising sonority then various phonological changes occur so that the output has equal or falling sonority over the syllable boundary. This is exemplified by underlying bimorphemic /kuk-mul/ which is realized as [kuŋ.mul] ‘broth’ and /kam-li/ which is realized as [kam.ni] ‘supervision’. Both inputs have a rising sonority medial cluster but both output clusters are equal sonority. Thus, Korean witnesses an asymmetry in syllable contact sequences: rising sonority clusters are strictly avoided while falling and equal sonority clusters are normally allowed.
In general, I don't think it is too unusual for a language to disallow or disprefer obstruent + sonorant sequences assuming that it doesn't permit these sequences to be syllabified as an onset. Permitting these onsets is common (which would explain why these sequences are common cross-linguistically), but in a language that doesn't permit them but does permit coda obstruents, we would have a coda obstruent followed by a more sonorous onset, which might not be optimal for the perception of syllable boundaries, and which I think wouldn't qualify as "the least marked kind of cluster". The development of gemination in medial -Cj-, -Cr-, or -Cl- clusters in some languages (e.g. in some contexts in Germanic or Italian) has sometimes been attributed to a tendency to avoid heterosyllabic -C.j-, -C.r-, -C.l- sequences.
Some languages don't permit obstruent + sonorant sequences but also don't permit sequences of non-nasal sonorant + obstruent. E.g. Pali shows complete assimilation of sonorants in both obstruent + sonorant sequences (yielding geminates in medial position) and in sequences of non-nasal sonorant + obstruent. Japanese doesn't allow either obstruent + sonorant or non-nasal sonorant + obstruent.
I suspected that there might be Turkic languages that also forbid obstruent + sonorant sequences (since from what I recall, there is a tendency for Turkic languages to not allow complex onsets), and it seems Kazakh might fit this criterion:
Kazakh does not permit onset clusters; therefore, no Kazakh words start with a sequence of two consonants. [...] Kazakh does, however, have coda clusters, although there are significant conditions on their distribution. The most important one of these is that while the first member of the cluster needs to be a sonorant, i.e. /m, n, ŋ, r, l, w, j/, the second member has to be an obstruent, meaning that Kazakh strictly abides by the Sonority Sequencing Principle [...] changes in the quality of Kazakh consonants occur also because of reasons related directly to markedness constraints targeting coda/onset syllabification. One such constraint is the Syllable Contact Law (SCL), which Kazakh strictly abides by, unlike many other Turkic languages. As per the SCL, an onset consonant in Kazakh has to be less sonorous (or at least cannot be more sonorous) than the preceding coda consonant.
("Kazakh phonology", Özçelik, Öner. To Appear in the Encyclopedia of Turkic Languages. Brill.)
Alternations in suffixed words suggest that sonorants in Kazakh have undergone fortition to stops in this position.
Kazakh is also discussed in Stuart Davis's "Syllable Contact in Optimality Theory".
See also this paper about Kyrgyz, which has suffix alternations that show avoidance of obstruent + /n/ or /l/, although it allows sequences of obstruent + /m/: "Sonorant Restrictions in Kyrgyz ," Hanzhi Zhu
Zhu mentions that this may be correlated with restrictions on word-initial sonorants, and it seems that Proto-Uralic possibly did not allow word-initial *r.
