Iām noticing some unusual inconsistencies in how scripts execute on databricks.com compared to when the same workflow is triggered through a mobile-based API. On Databricks, the script runs perfectly when executed directly inside a cluster notebook. But when triggered through an API call, the execution timing and even some output behaviors change slightly ā which ends up affecting downstream tasks. To check whether this is a Databricks-specific behavior or a broader runtime/environment issue, I started comparing execution across other platforms as well. I even tested how lightweight mobile script executors handle similar runtime variations (for example, tools like Delta Executor APK ). Surprisingly, the pattern of environment-dependent execution differences appears in multiple platforms. So my question to the community is: What typically causes scripts to behave differently between direct cluster execution on databricks.com and API-triggered runs? Could it be: Environment variables? Session initialization? Cluster warm-up? API gateway timeout differences? Or something else affecting runtime consistency? Any insights will help a lot ā trying to determine if this is a Databricks-side factor or a universal runtime behavior issue.
Ellie
