Abstract: Techniques are described for “fuzz testing” serverless software applications (for example, software applications implemented by a serverless compute service provided by a service provider network). Fuzz testing is an automated testing technique involving a fuzzer client application (or simply “fuzzer”) providing invalid, unexpected, or random data as inputs to a software application that is the target of the testing to discover software bugs and other runtime issues during execution of the software. To fuzz test a serverless application, a fuzzer continuously sends inputs to the serverless application and, when a runtime error is detected (for example, by monitoring log data generated by the serverless application), a copy of the serverless application can be invoked with the suspected error-inducing input to verify that the input indeed causes the error without interrupting the fuzzer's testing of the original serverless application.