On the velar nasal /ŋ/ sound followed by /k/
Solution 1:
As mentioned in the comments, this previous question has overlap with yours: Why do dictionaries transcribe the nasal in 'think' and 'language' with /ŋ/, yet 'input' and 'inbox' with /n/, not /m/? I don't want to copy my answer there, so please go to the linked page to read it.
To address your numbered questions:
-
According to many accounts, English speakers may optionally pronounce coda /n/ as [ŋ] before any syllable starting with velar plosive, even when there is a word boundary. So in this way, [ŋ] would be possible not only in words like income, but even in phrases like "in ten cars".
However, I've read that the "[ŋ]" produced from /n/ by a process of "gestural overlap" like this may actually be different, either acoustically or maybe just articulatorily, from the [ŋ] sound used for the phoneme /ŋ/.
This kind of gestural overlap across syllable boundaries does not affect the phonemic status of /n/. For example, I have /n/ and not /ŋ/ in the word nightingale, which is detectible from my pronunciation of the /t/: it's a voiceless stop for me, as in the word lighten, rather than being lenited and voiced as in the word lighting.
The same kind of gestural overlap is supposed to cause /n/ to be pronounced like [m] before labial stops or the labial nasal, /t/ to be pronounced like [p] before labials and like [k] before velars, and /d/ to be pronounced like [b] before labials and like [g] before velars.
I don't think information about these kinds of gestural overlap is useful to ESL students.
-
For derived words, refer to the pronunciation of the original word. For example, clinking, banker, linkage have /ŋ/ just as clink, bank, link have /ŋ/.
It's probably also true to say that /nk/ is impossible as a syllable-final cluster (that is to say, /n.k/ only occurs with an intervening syllable boundary), but the problem is that English syllable boundaries are difficult to place and people disagree about where they fall.
For non-derived words, intervocalic "nk" is probably /ŋk/, although there might be exceptions.
I don't know the actual etymology of the word inkling, but it looks like it ends in either a diminutive ending -ling or a frequentive suffix -le followed by the suffix -ing. Removing -ling gets you ink, which would have to have /ŋk/ because, as you mentioned, /nk/ is not possible word-finally.
-
(also 4) The distinction between /n/ and /ŋ/ at the end of the prefixes in- or con- before a velar consonant, as in the word conclude, is not too important to native speakers. I don't know the figures about how well people can distinguish the sounds in this context.
I think every native speaker would agree that /n/ and not /ŋ/ is the phoneme that shows up in compound words like pancake or pincushion.