Using complex collision as simple uses a vertex-for-vertex recreation of the asset as the collision data. The disadvantage to so doing is that it’s an increased performance hit (it effectively has to draw the sword twice) though unless it has a stupid-high vertex count it’s probably not a big deal; more of an issue for geometry that’s iterated a lot across a level (e.g. using complex collision as simple for assets like walls and doors can chew up GPU power pretty quick since that’s effectively doubling your whole level geometry).
The real disadvantage, imo, is you didn’t learn how to solve the problem! By doing what you did you essentially changed the size of the collision geometry in a way that relegated your weird bugs with collision to edge cases because the sword is much less likely to strike the player capsule. But they’ll still arise; it’s better to read the docs on collision channels and learn to set object/component collision settings since you’ll have to do so eventually anyway.