layer 2 development is sold as a speed and cost fix. and it is. but the conversation nobody has upfront is about sequencer design, data availability, and what happens when you need to settle back to L1 at scale. clients come to us wanting "a layer 2." by the end of the first call, they're asking questions they didn't know existed a week ago. that's not a problem with the client. that's just how deep this actually goes. the projects that do it well spend real time on the architecture before a single line gets written. the ones that rush it end up rebuilding the whole thing six months later on someone else's budget.