Clones in cEDH

One Clone, Two Clone, Red Clone, Blue Clone
Clones are a bit of an oddball mechanic in Commander, breaking the traditional singleton environment and offering a taste of what it feels like to play with multiples of a card. Similar to tutors, clones also bring a unique flexibility in their play patterns: they may come with a restriction or two, but they are far from fixed duplicates (at least, in theory). That being said, a clone is only as good as its doppelgänger, and with common mana costs ranging from the fragile Phantasmal Image
I won't waste time burying the lede any deeper: whether or not you should be dedicating slots to clones comes down to roughly two questions. First off, in the ideal case, what would you like to clone? Secondly, if you're able to one consistently clone your targets of choice, how does that compare against removal and/or stack interaction for that target? In essence, is the clone slot better dedicated to stopping a threat from happening in the first place? Or is the flexibility and value offered so great and/or unique that the correct interaction is to simply double down?
That's a lot of deckbuilding introspection, so let's take things one at a time by splitting up clones into two categories: cloning your own things and cloning your opponents' things.
In-Clone
When it comes to clone evaluating, starting with cloning your own creatures helps to narrow things down a bit because we can essentially ignore the question of removal/stack interaction. Clones used for this purpose certainly aren't going to be replacing slots which would otherwise answer a threat; I mean, bolting your own Krark, the Thumbless
This leaves us to the other half of the equation: the value proposition. While self-cloning brings with it reliability in terms of strategy (we know our deck and can play so as to maximize the usefulness of a well-timed clone), we do sacrifice some value due to the issue of dependency. A clone used to double up on our one resources is only as good as what we already have in play, meaning that they can often serve as unfortunate "win-more" cards if not built around properly. Drawing a clone on an empty board is effectively a missed draw for the turn, and nothing hurts more than a wasted card slot. Conversely, drawing a clone after consolidating a board presence can frequently leapfrog you ahead of the rest of the table, creating a dominating position which is difficult to eclipse.
Ultimately, this serves to highlight the gamble which comes with self-targeted clones. When they are good, they are very good, and when they are bad, chances are that they will be useless. Their real value lies in the context of the rest of your deck: any opportunity to leverage list construction so as to minimize the chances of an empty board state will directly impact the viability of your clone package. If your standard target is your commander, all the better. Odds are you'll see that creature at least once in your game.
Speaking of commanders, it's worth directly mentioning the four cards which can copy legendary creatures, setting them apart from other clones and firmly in the "in-clone" category. They are: Sakashima of a Thousand Faces
First up, the Sakashimas. Not much more can be said about this pair beyond what we've already discussed for clones, except of course for the fact that good ol' Sakashima of a Thousand Faces
Next up, Spark Double
Finally, we come to Irenicus's Vile Duplication
Out-Clone
Our second class of clones complicates things a bit more, as now we have to weigh their inclusion against that of stack interaction or removal. Let's work backwards in terms of severity, starting with an all-or-nothing case.
You are Player A, and it's Player B's turn. Player B casts a Dockside Extortionist
Rewinding this, and let's swap out the Phantasmal Image
While this example may seem a bit absurd, it actually teaches us a lot about the fundamentals of clones. Clones are as good as your opponents' board states, meaning that they can help you jump forward to catch up, but the better your opponents are doing then odds are their creatures might outpace your ability to clone them. Cloning a Dockside Extortionist
Let's look at another out-clone example, this time not quite as all-or-nothing.
Once again, you are player A, but this time Player B is deploying a Kraum, Ludevic's Opus
In this case, your clone serves as a statement that, rather than deny one player resources, you would prefer to advance yourself so as to combat each of your opponents more directly. This may sound a bit strange, but let's think about it. Had you stopped Kraum, Player B would likely have been put into an irrecoverable position. However, by waiting a full turn cycle to copy Kraum, you've provided Player B with an opportunity to take several game actions, potentially draw several cards off of Kraum triggers, only for you to eventually take advantage of a cloned engine which now needs to split its card-draw offering to hold off your three enemies.
The second case certainly sounds a bit dower - and don't get me wrong, it certainly can be in a lot of situations - but what it does provide for is excellent politicking. By keeping Player B in the game, as opposed to removing or countering Kraum, Ludevic's Opus
When it comes to deciding on what out-clones to run, the question is almost entirely one of efficiency. Magic has a thirty-year history of printing clones, so it's only a matter of the cheapest to pick from. Phantasmal Image
Wrap Up
Clones are a funny beast: they may be the archetype of duplication, but for such a many-faced creature, they seem to have taken two distinct paths. Whether you're doubling down on your own creatures or sneaking a copy of your opponents', one thing is for sure: clones are here to stay, we just need to figure out where they'll be happiest in a cEDH home.