It is grammatically incorrect, but most people think it is natural and don't care.
The simple subject is thing, which is singular, so the verb must be singular to agree with it. Notice that in the subordinate clause, scares is singular, agreeing with thing.
The speaker got confused because it seems strange to say that a singular thing "is" a plural like oranges. That's more than one orange, right? Perhaps the speaker could revise the sentence so the subject is "the things that scare me the most". But there are two problems with that. First, the idiomatic phrase is "the thing that scares me the most". If you say "the things that scare me the most", that's weaker, because then you aren't pointing out the #1 scariest thing. Second, the speaker starts saying the sentence before choosing the main verb. By the time the speaker has said "The thing that scares me the most", it's too late to change it. At that point, the speaker sees oranges coming up ahead, and scrambles to choose the verb. "Is" sounds wrong because of oranges; "are" sounds wrong because of thing. Oh no!! There's no time to think this through, so the speaker compromises and chooses a verb to agree with oranges, dimly sensing that something is wrong.
And then life goes on.
There is no reason to think that "audience" is being treated as plural in the sentence, so why would you object to a / an? (I assume you've come across "a family", "a group", "a class", "a school", "a government", "a team".)
However, even when collective nouns are treated as plural, they can still take the indefinite article a / an. We could say, for example, "when an audience are happy with a performance, they often applaud", "when a government are disunited, the opposition often benefits", or "when a couple are fighting, their children often notice".
From Kevin Bridges in The Guardian, 15th October 2014:
When an audience are in raptures at a part of the show, the fear of the next bit not being as funny or causing a dip will set in but you can navigate that.
From Clare Dyson, AusDance, 2008:
Will an audience understand that they are voyeurs if they watch through peepholes?
Jodie Helm quoted in Caroline Heim's Audience As Performer, 2016:
Usher Jodie Payne knows when an audience "are really fragile, angry, enjoying themselves, visibly upset or offended".
Is your misconception that collective nouns can't take the indefinite article perhaps based on a confusion between collective nouns and mass nouns? These are two different things. "Audience" is almost always a countable noun (and according to Oxford, the use of "audience" as a mass noun is archaic). It is true that mass nouns can't take the indefinite article (although many mass nouns can be treated as countable under certain circumstances, often with a slightly different meaning - and when treated as countable, they can take the indefinite article - e.g. "a cheese" meaning "a type of cheese").
Best Answer
"Of the audience" is a prepositional phrase and can essentially be struck from the sentence to look at the rest of the structure. So look at:
Here it it obvious that the phrase should be "Most are..."