Having come across this term here for the first time, I wonder what it actually decribes. Is it a (pejorative) umbrella term for "conventional" rpgs as opposed to "storytelling"/"narrative-driven" games? That's at least what I've deduced.
It is indeed a new term used to describe conventional, normal RPGs in the "traditional" tabletop RPG format as opposed to newfangled indie games. It is not pejorative in nature, though it is used a little grudgingly as it mainly exists to distinguish "games that work like most every RPG ever as opposed to whatever crazy new variation you've come up with" in Internet discussions.
A trad game likely has:
D&D, GURPS, Rolemaster, and the vast majority of games published before the year 2000 are trad games. (Notable exception: Amber Diceless Roleplay). There are many new "trad" games too, from Savage Worlds to Eclipse Phase, that bring new genres or systems to gaming but stay within the traditional tabletop RPG format.
There is no clear legalistic differentiation between trad and indie. Most would say that the White Wolf Storyteller system, though it had initial aspirations to being narrative, is in retrospect a completely trad system. However, many new games have some aspects of traditional RPGs but innovate in one or a couple ways - there's no real clear "this crosses the line" criteria, it's more an attribute of self identification by the game's creator(s).
The way Trad has been used, it generally refers to class-based, class & level, random character generation, dice-driven mechanics, with a single GM with total rules and narrative authority. The only player input is what the characters attempt.
It's much easier to list, however, what makes a game not-trad:
Not definitive to trad nor non-trad
Some traits aren't themselves enough to toss something out of trad by themselves, but in combination, may be enough to do so:
You can identify a "traditional" game by looking at or inferring how its rules were developed. Were they developed by following tried-and-true traditions of RPG design, or were they developed by a different process? This is a way to identify "trad" vs. other games that is complementary with other ways that focus more on the rules features a game has or doesn't have, and looking at the background for why there are different ways of approaching rules development explains some of the fuzziness of the meaning of "trad".
In the beginning…
This method works because it's how the division came about, historically. First, there was the original Dungeons & Dragons. It was wildly successful, and it had many imitators. Some of the imitators were very good. Notably though, for many years most games that came out innovated by changing the details of how you handled something during play without changing what you handled – rolling percentile dice against a chart instead of a d20, adding skill systems instead of using GM judgement for what you could do outside combat, different ways of creating characters, how characters progressed or earned progression, etc.
Traditional games, then, are built using traditions. This is arguably a good thing, as it ensures that the basic structure of play will be familiar to anyone picking up a new game, and their experience with "roleplaying games" will be transitive to this new game. It means that the structure is tried-and-true, that the game will work more or less because it's following a successful design pattern. The innovative details have a good chance of succeeding because the game doesn't mess with the formula of "roleplaying game", giving the new mechanic a chance to shine against a background that's sturdy and solid.
Then people argued over what was good
Eventually, people argued about what games were best and there was much contention. But out of this contention came a vocabulary for describing why people found certain games satisfying and found certain other games unsatisfying, without having to label things as "bad" or "good". (For more reading on early theory, this set of posts on the emergence and content of RFGA theory is excellent.) It was the very beginning of creation a critical theory of roleplaying games (not to be confused with a scientific theory; think more like film or music theory), and it suddenly gave people new tools for understanding the structure of roleplaying games in their play and in their rules. Eventually, people starting using this new understanding – more or less consciously, depending on who it was – while writing new RPGs.
Suddenly, people were crafting rules not to be more realistic when taking damage or to provide a smoother XP curve; now people were crafting rules in order to do things that the traditional design never provided rules or structure for. Things like structuring the game like an action movie (Theatrix) instead of a simulation game, for just one illustrative example of the shift in thinking. Suddenly, there was a divergent evolutionary branch along which roleplaying games design could progress. Meanwhile, traditional games kept going, kept being made by iterating slowly on a proven framework of traditional design elements.
Even more eventually, there were refinements, disagreements, and new schools of RPG theory. Games were made to test, prove, or simply enjoy parts of these theoretical critical frameworks. These games were more or less (usually less) popular, but the people who made and shared them were getting out of them more of what they were looking for. The birth of theory, non-trad games, and the validation of a plethora of playing styles happened all in a tangle together. Of course, there was also a lot of bickering, One True Way–ism, and crowing about which way of playing RPGs was "superior". Those were fortunately the minority; though they were a loud minority, causing a lot of rifts with "traditional" players.
So now we are at today. We have two strong traditions of RPG design, one which is called "traditional" and one which is called nothing unified but for the purpose of this answer I'll call "theory-based", but they can both be thought of as traditions. Theory-based design is often more experimental, but it's had a decade or two to mature, so it has begun to result in games that are much more solidly-built and more widely successful than early games. They're not a unified whole though, hence the lack of a handy name for them: some are storygames, but lots aren't; some are independently published, but not all by far. They're only lumped together because they're such a small part of the pie chart – you know, that tiny slice that always gets the label "other".
These distinctions will change
Though the theory-based type of design was originally in reaction to the one-size-fits-all of trad games, since then there has been significant cross-pollination. D&D 4e's designers have been quoted as saying storygames had a significant impact on its design. You can see this in how focused the game is on doing one thing and doing it really well: where 3rd edition tried to be all things to everybody, accommodating plot-centric stories, character-driven drama, tactical combat, etcetera ad nauseum, 4th edition aimed to be a damned fine tactical combat and character-optimisation engine. Such a laser-tight focus on being a particular kind of game is characteristic of non-trad games, yet the sheer weight of tradition that D&D 4e carries and expresses in its structure makes it still very much a "trad" game.
This cross-pollination is to be expected though, since the bits of theory people have come up with are slowly, osmotically dispersing into the wider community of designers and players. Designers (moreso than players) are like musicians – good musicians don't listen to only their own music, or even only music in their own genre. They listen widely, and incorporate and play with themes and techniques from a diverse section of musical genres. Similarly, good designers don't only play their own games, because that would be stunting the growth of their design skills.
Once, the idea that you could classify motives for making in-play decisions into mixes of game-, drama-, or simulation-centric priorities (GDS theory aka the Threefold Model), or that you could categorise games by the proportions of gamist, narrativist, and simulationist leanings its rules had (a mutation of GNS theory) was weird and fringe thinking. Today these ideas are so pervasive that they're often taken for granted as true. This blending of the two traditions means that what is and isn't "trad" is a moving target as the strong "trad" trunk of RPG design slowly assimilates the most useful things from non-trad games the "trad" designers play.
And of course, there were always non-trad elements in trad games, even before the development of any conscious theories of RPGs. If there hadn't been, there would never have been elements that the early theorist could point to and say, "I like this, while I don't like that." Early games with non-trad elements like Amber Diceless Roleplaying blur the lines and give the lie to a strong distinction. Therefore it's easier to think of games that were designed with different design methodologies than to look at individual features of games – at least, you get a slightly clearer division that way.
Now (or always has been) more of a spectrum
But really, because of those older games with untraditional mutations, and because of the more recent cross-pollination, it will always be more sensible to think of games as scattered across a 2D (maybe 3D, actually) chart, and to recognise that there are clusters rather than distinct categories. D&D 4e is definitely in the trad cluster, though it's nearer the edge of that cloud of games than some others it shares company with. A game like the Burning Wheel is very solidly a non-trad game (and one of the few that are clearly in the "indie" cluster due to how it's published), but it draws so strongly on what the author liked in D&D campaigns that it's nearer to the trad cluster than, say, Microscope is.
And, as you can see, looking at it this way there is no inherent pejorative meaning to "trad", even if it can used as one by the intention of the speaker. Non-trad games arose in contention and dispute – which makes sense, because why invent something new when you're satisfied? – and there will likely always be a hint of discord in how people discuss different games, at least so long as there remains a recognisable gap that defines the "trad" cluster and the clusters of other games.