This is going to be a pretty dissatisfactory answer, but given that there hasn’t been another ventured in 3.5 months, I thought I’d share what I know.

For why it would be more common in moyen-français, I rather like this line from the historical summary of the period that you cited:

Le français perdit la prérogative de se développer librement, il devint la chose des lettrés, des poètes et des grammairiens.

This process accelerated with classicism. Perhaps spoken language would be less affected, but the written record certainly swerved towards regularization after that period.

This OQLF article implies that dissimilation is alive and well today, however, even if the spelling does not reflect the pronunciation.

Certainly both assimilation and dissimilation are active in English, as the English Wikipedia articles make clear, despite spelling standardizations that would hide it in the written records. But the written record doesn’t affect the acoustic and articulatory processes that underlie phonology.

I think there would likely be the same underlying cause for harer and colonel. I don’t believe nasalization or phonoation would play a role in either, only the acoustic and articulatory character of [r] and [l]. Whether it’s because it would be hard to pronounce two in a row or because it would be hard to distinguish two in a row is likely immaterial and difficult to establish for a particular centuries-old case anyhow, though you can sometimes guess at differentiation in modern cases. (One way in which dissimilation is not “just the reverse” of assimilation is that assimilation probably doesn’t have acoustic origins, i.e. it’s not done to confuse the listener!)

It doesn’t look like the final “r” was silent as early as harer was borrowed, but one reason they might have mentioned the future tense in particular is that in harera, etc. the second syllable would have lost the stress and that reduction brings the two consonants still closer in the articulation (as well as moving the acoustic salience that would have possibly impeded the shift — comparing it to the similar case of English governor as “govenor”, if you emphasize the second instead of the first syllable you can’t possibly delete the “r”).

Incidentally, one reason the change in colonel might have taken place in French and Spanish but not in Italian is that the second “l” was and is doubled in Italian. In Italian, geminates are phonemic and hence presumably already sufficiently distinct from single “l”.

More I can’t say, e.g. on dissimilation in moyen-français, since I don’t know. I do have contact with a historical linguistics professor focusing on Romance languages of the period, so if I have the occasion to ask her opinion I will. 🙂

‘The long-distance acoustic effect’, selon moi, doit être compris comme la présence soutenue d’un son dans la chaîne parlée et de son incidence sur l’articulation des sons voisins, probablement due au repositionnement de l’appareil vocal, qui est plus lent que la prononciation.

Je pense par exemple au mot inventer, dans lequel le V, de par sa position entre deux nasales, n’est pas exactement comme celui du mot valeur (je sens encore un filet d’air passer par le nez lorsque je le prononce, contrairement à celui de valeur). En fait, même pour les mots vent ou vingt, ma prononciation du V initial est déjà altérée par ma préparation à la voyelle nasale qui suit.

Concernant la dissimilation, si j’ai bien compris la notion (?), j’aurais quelques exemples (qui concernent cependant tous un unique phonème).

Souvent, en français du Québec:

• Enfant devient /ɑ̃fã/, et
  En enfantant devient /ɑ̃nɑ̃fɑ̃tã/.
• Cent se dit /sã/, mais
  Cent ans si dit /sɑ̃tã/.
• Entendre /ɑ̃tɑ̃dʁ(ə)/
  J’entends /ʒɑ̃tã/
  Nous entendons /nuzɑ̃tɑ̃dɔ̃/
  Malentendant /malɑ̃tɑ̃dã/.

Il est remarquable que le phonème [ɑ̃] subisse ainsi une déformation qui n’est pas fonction du mot en absolu, mais plutôt de son voisinage.

C’est comme si l’on cherchait à distinguer le dernier son [ɑ̃] d’un mot ou d’une expression afin de donner un petit signe à l’interlocuteur que l’élocution du concept (je dis concept à défaut d’un meilleur mot) est terminée et que son cerveau peut l’analyser comme un tout terminé.

Je me demande si certaines conjugaison du verbe faire n’auraient pas subi ce genre de mutation par dissimilation, expliquant les prononciations [ə] de l’imparfait (elle faisait [ɛlfəzɛ]) et des futur et conditionnel présent, où même la graphie est modifiée par rapport à l’infinitif (nous ferons).