Adam Brown very reasonably compares so-called linking /j, w/ with linking /r/, asking and answering his own question.
Is it worth teaching linking /j, w/? Well, one benefit is to avoid an overabundance of "linking" glottal stops in such situations.
I assume you would say the same thing about linking /r/. Just as 'two evils' and 'two weevils' are not the same, so Mark Anthony did not say "Friends, Romans, countrymen, lend me your rears."
Yes, and likewise more ice mɔːr aɪs sounds different from more rice mɔː raɪs — in the same way as an aim ən eɪm sounds different from a name ə neɪm. Syllable-final sonorants are shorter, more lightly pronounced, than the longer, more deliberate, syllable-initial ones.
The difference between linking r and the postulated linking semivowels is that native speakers (of RP-type accents) actually do typically insert an extra segment in the linking-r position. In non-rhotic accents the word more, pronounced in isolation, has a straightforward CV structure, m plus ɔː. Before a word beginning with a vowel, on the other hand, it can contain an extra segment, CVC, m plus ɔː plus r.
The analogy for my would be that in isolation it is CV, m plus aɪ, but in prevocalic position (my arms) it becomes CVC, m plus aɪ plus j. Since English j by definition is a non-syllabic palatal glide, and since aɪ already contains a non-syllabic palatal glide, it is difficult to see what the realization of j might actually consist of in the supposed homosyllabic sequence maɪj. (The case of I yearn or “my yarm” is different, with the palatal articulation now partly in the second syllable.) See vp’s comment on yesterday’s blog made early this morning.
I am confirmed in my view that “linking /j/ and /w/” are figments of the imagination. That does not necessarily imply that they are pedagogically valueless. I am willing to recognize that sometimes teaching something that is not strictly true may nevertheless be justified if it leads to better results than not teaching it. (That is the justification for some of the nonsense I encountered from the voice teacher whose classes on Accents and Dialects I attended anonymously when researching my own book on the subject — telling people to “place” the voice in the middle of the forehead, and suchlike.)
If talk of linking semivowels is a useful technique for getting EFL students to avoid prevocalic glottal stops (“hard attack”), then so be it. That might justify its use by teachers of learners whose first language is, for example, German.
But not of those whose first language is Spanish or Italian.
As for the claim that it is useful to help learners “avoid strong glottal attacks that can cause vocal issues”, I feel I need to see some kind of evidence that glottal stops cause voice problems. If the use of glottal stops really does trigger pathological conditions, you have to wonder how monoglot native speakers of (northern) German manage. In that language all initial vowels normally have a reinforcing glottal stop (“hard attack”). Given that, how do the Germans ever manage to remain in vocal health?