Here is summary of my investigations into ways to improve stability for the cicd by moving to an alternative infra then the current apache jenkins infra.
1) Alternative build script for apache jenkins This experiment was not successful, jenkins does not seem to pick the updated job config in all instances, I have reverted all jobs to there prior config. 2) Github CICD experimentation We had success using it and a pr is ready, but the service is not in GA and not available to all users. So it's currently not an option. 3) Azure pipeline https://github.com/aahmed-se/pulsar/pull/10 Tried with a forked repository things work well, there is even test report generation, apache infra won't allow integration since it requires an admin token to work. 4) Git mirroring This is strategy being pursued by projects like apache flink, they an external bot being managed by them that mirrors github pull requests to another repo which they grant admin access to ci cd services. They interact with it through phrases in the pr comments. 5) Appveyor and travis I have not tried these services but apache repos can interact with and it is supported by apache infra as they require a read only token. Travis instances are less stable so Appveyor is the better choice. It's unlikely it has the same capacity as azure pipelines. Conclusion The best and shortest course of action will be ask apache infra to enable appveyor integration and move the ci cd jobs there. It's already being used some apache projects like calcite and it should be sufficient for current use cases. Long term integration with github actions or Azure pipelines will be the best course of action.