We searched an effect monad to replace scala.concurrent.Future and move our tech stack away from 'Scala as better Java'.

The requirements for the replacement were:

• Battle-tested & documented;
• Type-safe;
• Fiber-safe (not leaking concurrent resources).

ZIO was picked for the following reasons:

• ZLayers (with zio-magic) that enabled us to structure the application & do DI in a functional way;
• Typed error channel for safer handling of business-logic exceptions;
• Built-in Scheduler and retries / cancellation;
• Steeper learning curve when compared to cats-effect.