Why does "singer" have /ŋ/ and "longer" have /ŋg/?

Singer and longer both end in the letters -er, but they don't end in the same suffix: singer ends in the -er suffix that forms agent nouns, while longer ends in the -er suffix that forms comparative adjectives.

This is the reason why I included the word "certain" in "the addition of certain suffixes".

Words ending in the comparative suffix -er or the superlative suffix -est could be said to be an exception as a class to the general pattern of suffixes not altering the pronunciation of [ŋ].

But there are very few words with the relevant sequence -ng- + comparative -er, since there aren't many monosyllabic adjectives ending in [ŋ]. In fact, there are only three adjectives that have inflected forms with [ŋg]: younger/youngest, stronger/strongest, longer/longest (all common). The adjective wrong, also common, has what I think are uncommon inflected forms that are pronounced according to dictionaries and my own intuition with [ŋ]: wronger, wrongest. So it's arguable whether there is a regular exception for the category of comparative and superlative forms, or it's just a matter of these specific three adjectives having special comparative and superlative forms.

The same distribution of [ŋ] and [ŋg] in adjective forms is described (and prescribed) by John Walker in his Critical Pronouncing Dictionary (1803) so it seems to have been around in its usual modern form for a least a couple of centuries. But Walker also mentions variation in his time, saying that in Ireland comparative adjective forms like longer are "generally pronounced" with [ŋ] rather than [ŋg] (lxxxi).

In informal off-the-cuff formations, you can sometimes find comparatives or superlatives formed from words of two or more syllables ending in -ing, which I believe like wronger, wrongest would not be pronounced with [ŋg], but with [ŋ]: things like charmingest and boringer.

When you know that the agent noun "one that longs" (longer) is pronounced without a /ɡ/: /lɒŋ.ə/ (Longman Pronunciation Dictionary, RP), this might be explained cursorily by referring to the fickleness of usage .

My daughter who attended the British School of Paris was reprimanded for saying ‘singing’ with a soft g sound, her teacher insisting on sin ng ing. Fortunately my daughter had the presence of mind to respond that in « Singing in the Rain » they got it wrong.