An old question, but one which I feel is worth answering for the benefit of other people interested in Semitic etymologies. The answer below me is excellent in stating that the root-and-pattern system enables speakers to more easily recognize allophones, but probably the most important contribution of root-and-pattern morphology to consonant evolution is that it provides a large number of potential candidates for analogical restorations and extensions. I'll be drawing on data from Biblical Hebrew, since that's the Semitic language whose linear evolution I understand the best.
In Proto-Hebrew, *n assimilates into a following consonant. From the root *n-t-n 'to give', for example, we get וַיִּתֵּן wayyitten 'and he gave', which historically comes from *wayyenten. But, in some verbs, an original *n is actually restored in positions where it should etymologically have been lost. Hebrew has restored *n by analogy between the forms verbs without any *n on the one side, and the forms of verbs with *n where the *n was not assimilted on the other. For example:
קָטַל qåṭal 'he killed' is to יִקְטֹל yiqṭol 'he will kill' as
נָקַב nåqav 'he pierced' is to יִנְקֹב yinqov 'he will
pierce' (the form יִקֹּב yiqqov, with assimilation of *n, is also found, reflecting the original, non-analogized form)
There are other examples of this kind of thing happening. The sequence *aʔ generally becomes *ō in closed syllables, ex. *wayyáʔmir → וַיֹּאמֶר wayyómɛr 'and he said'; but, sometimes, this same root shows up with the original *ʔ restored by analogy, ex.:
קָטַל qåṭal is to יִקְטֹל yiqṭol as אָמַר ʔåmar 'he said' is to
יֶאֱמֹר yɛʔɛ̆mor 'he will say' ([ɛʔɛ̆] is the synchronic outcome of /iʔ/
before a consonant.)
And a counterexample, where this kind of analogical leveling actually destabilizes a stable historical consonant instead of restoring an unstable one: *aw normally becomes *ō, but *w in word-initial positions becomes *y. So, for the root w-d-ʕ 'to know', the perfective evolves like *wadaʕa > *yadaʕa > יָדַע yåðaʕ 'he knew'. We would normally expect the imperfective to be יוֹדַע yoðaʕ from *yawdaʕu, but instead we get יֵדַע yeðaʕ, as if it were from *yaydaʕu. What has happened is an analogy. Since the perfective form begins in *y after the *w > *y shift, the imperfectve becomes constructed as if it were from a root beginning in *y as well, even though historically, *w should have retained its original pronunciation in this position.
So, overall: It's a mixed bag. There are certainly inbuilt paradigmatic mechanisms in Semitic languages which do promote consonant conservatism to a degree, but not absolutely, and these same mechanisms can just as well create new innovations as they create retentions. This can even lead to free variants and irregular patterns in which some portions of the lexicon have been affected by an analogical process whereas others were allowed to remain.