The Synoptic Problem
The Synoptic Problem may shed some light on this.
Matthew is also repetitive on this matter. He recounts this principle--or something very close to it--4 times (18:4, 19:30, 20:16, 20:27).
I have argued elsewhere that Mark exhibits an oral style, contra Matthew & Luke, which exhibit a more literary style. Put another way, Mark is written the way you would tend to speak a story. Matthew & Luke are written the way you would generally write a story.
This is supported by the (very probably independent) statements of Papias of Hieropolis & Clement of Alexandria that Mark wrote based on what Peter preached (see HE 2.15, 3.39, & 6.14) (for those who can't get enough of the Synoptic Problem and are interested in how putting Mark through an oral filter fits with other synoptic data, see my work here).
How Matthew presents this topic
The context in which Matthew teaches this principal is instructive: in each case, it accompanies a discussion where Jesus predicts His coming sacrifice. In fact, the 3 instances where Matthew teaches this concept using exactly the same word (πρῶτος) are all arrayed in a careful, chiastic structure:
- The first shall be last
- Parable illustrating this point (laborers in the vineyard)
- The first shall be last
- Jesus' personal example of this principle (He predicts His coming suffering)
- James & John's inappropriate (vain?) request
- Jesus' personal example contrasted with Gentile rulers (hey sons of Zebedee, which do you want to be like?)
- The first shall be the servant
Jesus teaches the principle, gives examples of applying it right & wrong in a story, reinforces the principle, gives another example of applying it right & wrong--this time in a concrete setting, and reinforces the principle once more with a memorable one-liner. The center (heart/most important part) of this chiasm: the first shall be last, as demonstrated by Jesus' sacrifice.
Matthew's only other presentation of this idea (18:4) is also in a passion-prediction setting, but he uses entirely different words--the teaching is conceptually similar but it's a loose parallel. The 3 really close parallels are all in the aforementioned chiasmus.
How Mark presents this topic
This opens an intriguing possibility--if Mark or Peter is working (at least partially) from memory, Matthew's presentation of this teaching can explains Mark's. Matthew's repetition of the last shall be first in chiasm, to highlight what Jesus is doing for the world, create a memorable association between predicting the passion & the last shall be first.
Example 1 from OP:
If Mark is working with Matthew & Luke as sources (see Synoptic Problem links above for why I believe that--whether you do or not let's run with it for a minute and consider its explanatory power), when Mark encounters the passion prediction in Matthew 17:22-23/Luke 9:44, he follows Luke's order (which he's doing at this point in his Gospel) and proceeds to sharing Luke's next teaching--the least shall be great. But instead of using Luke's words, Mark recounts the more memorable version of the idea found in later chapters of Matthew: the last shall be first. Mark/Peter has associated these 2 concepts (probably because Jesus associated them).
In the parallel passages, Matthew & Luke do not have the wording Mark uses in Mark 9:35, but Mark/Peter quoted from memory.
Examples 2 & 3 from OP:
In this passage, Mark is also presenting the material in a chiastic structure essentially identical to that of Matthew (though Mark's chiasm is smaller).
Mark actually really likes 3 part chiasms (technically it's a specific type of chiastic structure called an intercalation), in which he tells part of one story, blends it into another story, and then goes back to finish the first story. This specific story-telling methodology is often referred to as a "Markan sandwich". Mark did not invent the technique (it shows up in the OT), but he very prominently employed it.
Given the presence of chiastic teachings in the other Gospels (as well as in the OT), and the fact that different Gospels present material chiastically in different settings, I suggest putting Jesus' teachings into chiastic form is not an invention of any of the evangelists--more than likely, it was a method employed by Jesus Himself.
Thus, in examples 2 & 3 from the OP, Mark (like Matthew) preserves Jesus' teaching in chiastic form, hence the repetition.
Mark repeats this teaching in chapter 10 because it's part of a chiasmus in Matthew (shortened to a Markan Sandwich in Mark). The oral/memory hypothesis for Mark's origin offers an explanation for his usage of the same teaching in another setting in chapter 9--he associated the last shall be first with Jesus' predictions of His coming suffering, which is exactly how the concept is always presented in his primary source, Matthew.
Bonus material on the Synoptic Problem
Some may be shocked that I have set aside the axiom of Markan Priority in this post. I actually find this passage to be problematic for the popular Two-Source Hypothesis.
In Mark 9:35, Mark uses the words πρῶτος & ἔσχατος for "first" & "last", which Matthew does elsewhere, but in the passages that parallel Mark 9:35, neither Matthew nor Luke use either word. On the Two-Source hypothesis we must either conclude:
- Matthew & Luke both got this passage out of Q (which opens its own large can of worms) OR
- Matthew & Luke both independently made the choice to replace πρῶτος with μέγας. That this kind of fluke could happen once or twice by accident may be credible--but that it happened hundreds of times is quite difficult to believe.
This is a classic case of a minor agreement between Matthew & Luke that is most readily explained if Luke knew Matthew (or Matthew knew Luke).
If we're willing to make additional assumptions we could provide a similar explanation on the Farrer hypothesis, but I suggest that Markan posteriority (Mark wrote 3rd) offers the simplest solution to this textual anomaly.
Mark's reliance on memory does not require that he had memorized Matthew (or Luke). A scroll takes 2 hands to manipulate, so there was no way for one person to handle 3 scrolls at the same time. Even if Mark were working with two written sources that were readily available, it would be easier to quote sections from memory and periodically return to the source scrolls for reference, than to walk through the sources word for word.