To me, fanfic characterization revolves around the extrapolation of traits that we see in canon. The behaviors that a character exhibits -- what they say, what they do, how they behave with others -- are like points on a grid. They cluster together around certain ideas, and those are what we come to think of as a characters core traits. We draw circles and arrows and lines around those, and extrapolate where the end point might be if things continued along in a similar fashion, or we try to break up the pattern by deflecting the line and making it go a different way.

Do a good job of inference or show all your work in deflecting that line and the story will usually be thought of as 'in character.'

However, there are outliers to those core characteristics; behaviors and thoughts that seem really outside of what the character 'normally' does. Some fans go so far as to say those characteristics are OOC -- even if they are shown in canon -- because they are so far outside of what their expectations are based on how they have graphed the points in the characters past history.

And sometimes, the author will be working from only 3 points of data to draw their lines, or stick only to their 'preferred behaviors' even if those happened in only a couple of episodes. (This happens with Duncan and Methos all the time.) An author will take those 3 points and create a completely new behavior chart, extrapolating from that.

And sometimes, occasionally, when the author hasn't seen much canon, some of those points are crafted from stereotypes and fanon.

So for the people that see those three points as truly describing the character's core, those stories are recognizably in-character even without a lot of work. They can see the line, the path, where it started and where it goes.

But for those that see those three points as outliers -- well, they're screwed. They aren't on the same page as the author when the story starts, so there's no way they can get to the same place when the story ends.

And they are likely to be the ones that say "that story was out-of-character for me."

katallison's One for the Road, as is Something Borrowed, Something Blue, the Methos-as-serial-killer story.   

Both of these stories are internally consistent and highly recommended. They just don't have my Methos or Duncan in them.

The Methos-as-serial-killer one is actually easier to discuss, as it fiats in that Methos was not just tempted by Kronos (which he was) but that he has periodically indulged in killing sprees after leaving the horsemen (the non-canonical what if moment.)  The author's logic is simple: if he has done it in the past, he would do that now.

The bone of contention? Once he turned away from the slaughter, did he ever fall? Author says yes, canon, though, seems to indicate 'No.'  The theory here being that if he had said 'yes' to slaughter in the past, why not say yes to Kronos now, when Kronos is tempting him so?  And if he says no now, why does he say 'yes' two weeks from now or why did he say 'yes' two years ago? 

To write this story, the author has to ignore all of these questions, and go with 'well, sometimes he says yes.' Some people see that as a good enough response and the story works for them.  For others, myself included, it's not a good enough answer and so despite the internal consistency, the story falls flat.

Three points can create a line, or a plane, or a triangles -- and sometimes it creates a damn fine story. 

And sometimes it's just …three points.