I think the reason is concurrency model. Go is highly concurrent language in nature and core syntax. Scenario when some computation localized to concurrent process failure, but hole system continue to work can be viewed as normal. To my mind panic and recover are about failure handling, not exceptions.
