I’m saying the same thing as JJ Hantsch, but with different emphasis.
Is there a tendency for languages to gain or lose complex clusters over time? Both, but with different causes. Languages lose complex clusters through various processes of phonotactic simplification. Languages gain complex clusters through dropping vowels [EDIT: or adding consonants for ease of pronunciation: Epenthesis].
The gaining of clusters is pretty random, because unaccented vowels can drop out anywhere. Northern Greek dropped all unstressed high vowels, and all of a sudden had to deal with clusters like piðo > pðo “I jump” or mikros > mkros “small”. (At least some variants have dealt with it by epenthesis: imkros. Because mkros NCCVC is an unusual cluster, at least with the NC not being homorganic.)
The simplifications are not fully predictable: for example, Tsakonian has pummelled Greek clusters into CV, with unusual shifts like st > tʰ and tr > tʃ. But yes, I think we can say there is a tendency towards CV in language; it’s just a gradual, random, piecemeal, and easily disrupted tendency—to the extent that it’s not much of a tendency.
Do syllable templates change, as in languages going from CCCVCCC to CV? They can; Tsakonian did. And they can go the other way too, as Northern Greek did.
That’s why dialectology is awesome btw. You get to see multiple outcomes from the single starting point.