These are illustrative scenarios showing how the products are used, not customer case studies.
The same sequence, twice
Today
- Call the slow service
- Hold open or poll
- Status table drifts
- Restart loses the thread
With SynaptaGrid
- Step hands the work over
- Run parks with a deadline
- Service calls back
- Run resumes where it was
The situation
A rendering job, a credit decision, an overnight batch. Holding a connection open is not an option, so it becomes a polling loop and a status table — and a worker restart in the middle loses track of which requests are still outstanding.
How it is approached
The step hands the work over and parks. The external service posts its result back when it is done, authenticated with a token issued for that run, and the step carries a deadline so it cannot wait forever. Run state lives in the database rather than in a process, so a restart resumes the run, and steps are claimed so that running several workers never fires the same step twice.
What changes
A slow dependency stops shaping the architecture around it. A run that spans hours is watched, recovered and reported on exactly like one that takes seconds.