Rollups are an L2 solution that sits on top of a base layer. The idea is that these projects inherit many "security properties" from the base layer they use. Well, one of these properties or characteristics is that you cannot get trapped in L2 with native L1 assets secured by the bridge contract. How could this happen? In an extreme case, if the rollup sequencer disappears or fails, you no longer have a way to send tx to the base layer. That's simple.
For this reason, all rollups are asked to have a mechanism for forced exit at some point. This kind of mechanism consists of any user being able to return to the base layer without needing sequencers. It's not a normal transaction and perhaps more complex, but it's an exit option, and users aren't limited by the existence of a sequencer. Having this type of mechanism is one more item in a package of conditions that are currently being demanded of rollups to consider them a true rollup. Obviously, they're in their early stages, but it's expected that all will have it at some point.
An example of one that has it is dYdX, which is an "application-specific" zkrollup that works with StarkEx technology. Being an application-specific zkrollup makes it easier to design this kind of mechanism.
What's this mechanism called? It's called a "forced exit hatch." This means that users can exit L2 in case of an emergency, similar to a hatch that opens on a boat in case of need.
It's important to note that this mechanism doesn't guarantee total security, but it does help minimize risks. While there's still work to be done, the presence of forced exit mechanisms in rollups is an important step towards widespread adoption of this technology.
In conclusion, rollups are a promising L2 solution but still in development. The presence of a forced exit mechanism is an important feature for evaluating the security of a rollup.