NXP-25358 was about using AbstractBatchComputation to do this but as explained we don't want and don't need the batch feature at the action level.
But we still want to have a retry and failure policy per action so the behavior in case of problem is defined.
This could be done by having a generic AbstractRetryComputation that works like AbstractBatchComputation without the batching.
So can have ComputationPolicy that define the FailSafe rules for retry and what to do on failure: skip or stop.
The AbstractBulkComputation will simply extend this.
Also the BulkStatus should be extended to get an error flag and error message, so a command can be completed in error if the policy is "skip on error".