The risk involved is that you plan your moves assuming your L-cancel will go through; your entire strategy is based on L-canceling. It affects which attack you use when. So if you blow it, you're much more likely to be punished, especially if you opponent can predict that you'll miss it. Take Falco, for example: his entire metagame assumes that you'll never miss an L-cancel. If you do, he gets shieldgrabbed because his optimal playstyle is so aggressive, and Falco can't really afford to take many hits. That's why, when I fight Falcos, I generally try to lure them toward the edge, so that, when they miss the cancel, I can b-throw them off the stage, which is a very dangerous position. If L-canceling didn't exist, Falco players wouldn't approach. They would laser-spam constantly from a safe distance, and they'd be very annoying.
Or, suppose I'm playing G&W, who isn't fast and typically has a very predictable approach. I know that you know I'm going to SHFFL an f-air at you, so I know you're going to go for the shieldgrab and generally expect to punish me. That's why I'll throw out an f-air that stops just short, so you'll rush in to punish and I'll d-tilt or jab or whatever. But if I miss that L-cancel on the f-air, the entire strategy was worthless, and G&W becomes completely vulnerable/worthless.
Will I ever not want to L-cancel? No, of course not; the reward far outweighs the risk for a competent player. But if I blow it, I'm in serious trouble because of how I've planned my strategy. That's the risk.