Dependency Manager Performance Optimisation Enhancement

During Scottish Government (SG) analysis of long running instances of the dependency manager batch it was found one particular SQL query was 'expensive' in terms of time and didn't seem to be scaling well as cases and dependency records increased, albeit behaving as designed.

SG introduced a custom fix with support from product development (PD) to isolate precedent change set items that have no dependents (and hence, no case impacted) and running the dependency manager twice (once without and once with the no dependent items).

This has had a significant impact/improvement on the performance of the dependency manager and ideally should be considered for being taken into product.

Reducing the number of precedent change set item (PCSI) records in a precedent change set (PCS) would lessen the workload of this query and potentially many of the PCSI records in a PCS are 'irrelevant' / disconnected in that they have no dependents.

One of the questions asked by SG is, would it be possible to skip writing these disconnected PCSI records in the first place? Speaking with PD, we are aware that the existing current strategy is that every transaction likely creates a number of PCI records to reflect all the data changed but there is no filtering applied.

Based on further conversations with PD, a suggested option is to add a step to one of the OOTB dependency manager (DM) batches to identify and remove disconnected PCI records before existing DM processing continues. This is what this RFE is requesting.

Appreciate that an exact quantifiable improvement from this change is not known beforehand and more analysis would be required but as mentioned, SG have seen a significant performance improvement after making a similar local change.