I think it is important to look at the general tendency in which the language is evolving. If a language is loosing some distinctions, then there will be an important trend of speaking/writing without the distictions that are going archaic. There will be a counter-trend to keep them. And among such counter-trend, we will find hypercorrections, in which people will try to "keep" the distinction in contexts where the standard doesn't have it first place.
For instance, in Portuguese there is a general trend to lose verbal person desinences (except the singular first person and the plural third). There is also a tendency to change personal pronouns, especially the second person ones, and, more recently, and still less remarkably, the first person plural.
So, in a more concrete example, here is a Standard Portuguese sentence:
Nós estamos ficando velhos. (We are growing old.)
And this is like the same sentence will probably be rendered a few decades into the future:
A gente está ficando velho.
with the loss of the -mos desinence that marks the first person plural, and the change of the pronoun, from "nós" to "a gente".
Nowadays both sentences are already considered standard, though, a century ago, the second one would be considered wrong, or colloquial at best.
Now here is an intermediate sentence, in which the verbal ending is lost, but the pronoun remains unchanged:
Nós está ficando velho.
This sentence is quite unfortunate; it is too innovative for those who want to keep the standard, and too conservative for those who want it to change. Plus it poses a problem: "Nós" requires the predicative to go into the plural (nós... velhos). But the singular verb, "está", requires the predicative in the singular, so it is hard for speakers to decide whether it should be "nós está ficando velho" or "nós está ficando velhos". Probably due to such inconsistency, nowadays this sentence would be rejected as agrammatical by most speakers in both standard and colloquial registers. It can only be considered acceptable in the popular register - but then it has to get the verb changed, too: "nós tá ficando velho" (or "ficando véio").
And here is a "Piltdown fossil" of this evolutionary process, ie, a hypercorrection:
A gente estamos ficando velhos.
in which the dependent innovation (the pronoun change) is made, but not the innovation on which it depends (the verbal ending loss).
It is a common construction, but it has no future: the standard form of the language can accept the pronoun change, but not the incongruence between subject and verb; the popular register will avidly accept the pronoun change, but not the "sophisticated" verb flexion.
So here is a difference: if the language "evolves", then there are constructions that the standard register, that is most conservative, cannot accept, but become acceptable, or even the norm, in the more innovative registers, popular and colloquial (and will invade the literary register, too, since writers will feel the need to capture the gist of popular or colloquial speech to build their characters). But there are constructions that are misguided attempts to keep the formality of the standard, without really understanding the standard rules. Those can be quite widespread, but they represent neither the present nor the future of the language. In that sence, they are "wrong", not part of the evolution of the language.
Speaking of registers, it is necessary to notice that they have different grammatical requirements. What is acceptable in the popular register might be unnaceptable in the standard - but the converse is also true: some standard constructions are unnaceptable in popular register (I like to compare this to dress codes: you don't go to a night party dressed in a bath suit... but then you don't go to the beach in a tuxedo either).
When a construction is the standard norm, but becomes unacceptable in the colloquial or popular registers, it tends to go archaic. It will go archaic before it loses its social prestige. When a construction is the popular or colloquial norm, but is unacceptable in standard, its future depends on how much it represents a real trend of the language drift, or, on the contrary, how much it is a mere fad, destined to fade away. In any case, it will take more time to acquire prestige than to become widespread; when it acquires prestige, it will supercede the previous, now archaic, standard construction.
So, there are a few different things:
Pure mistakes, such as involuntary elisions, spoonerisms, typos, lexical misuses, etc.
Prestige mistakes - using popular constructions in formal contexts that require standard speech, or, conversely, standard constructions in contexts that demand popular speech.
Hypercorrections, in which the speaker tries to produce a standard construction, but due to insufficient knowledge of standard grammar, misapplies them.
Fads, usually slang terms, that become fashionable for a time, then become dated, rejected by most speakers, and finally go archaic, without ever acquiring standard status.
True innovation - neologisms, acronyms, portmanteaus, foreign loans that "catch", popular constructions that constitute the "natural" drift of the language, etc.
It is difficult to distinguish those things, even in one's first language. In a second language, it becomes almost impossible (so, rule of thumb, when speaking/writing a second language, it is best to err in the conservative side).
Hope this helps.
