I agree with sleemanj, in that using distributed strobes may be less prone to failures. There's just too much that could fail with having students keep 'beepers' on hand (dead batteries, dropped while playing basketball, left it in the bathroom, etc...). Since we can't assume an ability to distinguish colors/contrast well, perhaps a system of coded flashes would work better. This could be done with repetitive simple patterns, like 1x flash for fire, 2x flashes for earthquakes, 3x for lockdowns, etc. I imagine, if such systems become required by law, that is how they might be implemented (where single purpose "fire alarm" switches around campus are replaced with multi function switches next to them for other situations).
As far as implementation, ideally this could be integrated with existing fire strobe systems (which yes are a new requirement). The problem I see with that is you might be subject to the same red tape as fire system manufacturers (very drawn out/expensive process). So until laws are implemented that would grease those rails, you're probably best off making a standalone system.
There are many ways you could implement this. For a typical school I imagine using a wireless mesh system would be best... that way an alarm can be triggered from several locations, and the integrity of the system wouldn't depend on one device that could fail/be disabled. This could be done using something as simple as an embedded system like ESP8266 modules inside waterproof boxes with batteries. For more spread out/noisy environments, you could use 433MHz radios in a non-wifi embedded system (like a pro-mini or similar). If you wanted to only allow staff to trigger special alarms, you could make pocket sized triggering radios for that (this would make the strobe units much simpler to build waterproof).
All devices in the system should be designed so they self test periodically. One issue I imagine would come up is battery life/maintenance. Depending on how well it is designed, battery size, and how often the strobes are tested (presumably with monthly fire drills), this could become a significant expense. Adding solar could be a way to overcome this. Also it would be best to have a central monitoring node (like in the school office) that talks to all of the modules so it can alert staff of connection problems and low batteries.