(The following is inspired by a wonderful blog post from Mark R Johnson on the metagame of roguelikes.)
Let me begin by saying that the metagame is basically my favorite part of roguelikes - it was the gateway by which I entered the genre. It gives direction to the uninitiated, softens the blow of death, and bridges the gap between the charmingly-archaic roguelike core and modern triple-A "press button to engage content hose" design. It's an olive branch from roguelike designers to every person who ever booted up Spelunky and used all their bombs in the first thirty seconds. In the blog linked above, the author makes a very good and reasoned case against the metagame - specifically unlocks - which he claims weakens the core loop of die, analyze, regroup.
I understand why the author sees the metagame the way he does, but I contend that a slightly weakened core loop is preferable to a player disengaging from the game. Traditional roguelike design holds up accountability and discoverability over everything else. It eschews modern ideas of tutorialization and gentle learning curves; rather, you are pushed off the cliff and expected to learn how to fly on the way down. Along the way you'll see wax and feathers, but nothing telling you "hey, put this on there and then watch out for the sun!". The ideal outcome is that through trial and error, you eventually learn how to progress - but what if you don't? If you're like me, you end up hating the game pretty hard. Even players that don't eject from the experience can end up frustrated or even angry, which reduces their mental bandwidth, impeding their ability to focus and enjoy the game.
A specific grievance that the author levies against unlocks is that players generally end up polarized into two groups - his example cites FTL players; most are either unable to beat easy mode or able to regularly beat hard mode. The remaining players are the minority that can can only beat normal by using unlocked ships. The crux of his argument against unlocks appears to be that the people who are able to progress only with unlocked ships are not learning the correct lessons from death and have thus fallen out of the core loop. I posit that this narrow band of players is simply a transitional state for "competent" players en route to being "good" players, and that unlocks are the tool that allows them to make that leap. Without unlocks, that band is likely to be wholly subsumed into the "can't beat easy mode" tier.
A traditional roguelike gives you one way to progress your skills as a player: through purity of understanding and execution. A good metagame is an opportunity to provide players of disparate skill levels with the tools to make up for their shortcomings, to expose or hint at the wealth of tactics a beginner may not have considered, or to give the player a chance to practice a high risk / high reward combination. A good roguelike has breadth that the typical player may never be able to explore in its entirety - if that's the case, then what's the harm in giving them a few signposts to guide them? It's possible they still may disengage from the experience, but it's also possible that they become enriched by it.
Purists may insist that roguelikes aren't intended to be for everyone, but that just seems elitist to me. I make games for people to play; I want to know if someone has been pushed away from my game. I want to figure out why they got off track and give them a little nudge to get them back on. I think we all had training wheels on our bicycles as children, are we any less for it? Fortunately, the very nature of the roguelike genre means that the entire spectrum of challenge can be represented, and the presence of easy games does not dilute the appeal of hard games.