Text adventures are, almost by necessity, designed to railroad the player into doing something without it being obvious or annoying. Any mechanics introduced should be a natural extension to the world the player discovers, provide sufficient freedom so they don’t feel like a funnel, yet guide the player in the correct direction. In the text adventures I’ve listen to played on the Cortex/Upgrade crossover episodes, I’ve noticed that mechanics often act as hints. They are like bumpers on a wall (which may kill you, but you can make a different decision the next time.)
All these realizations have come after thinking carefully through the mechanics and puzzles I wrote which were clearly bad (and the few which were actually good), and trying to figure out what precisely caused them to have the effect on the players they did. It’s often me not paying attention to my good sense and the feedback of testers. More frequently it’s me trying to subdue the text adventure genre until it allows me to try and tell a story, rather than building a story that works within a framework.
Directly, most of my text adventure failures have come from a single mechanic undermining any positive decisions I made.
...