I ran a Hubbl scan and it looks like there are some 300k+ records under dlrs__LookupRollupSummaryScheduleItems__c object. We only have around 60 DLRSs. Does anyone know where these come from and how we can make sure we don't develop any tech debt?
there are two different paths here.
Scheduled/Watch for Change and Process Laterwill use the trigger to monitor DML and when a record might cause a recalculation, it will create a Scheduled Item record to enqueue a future recalculation. Needs the RollupJob Apex class scheduled to process those Item records.
Process Builder/Invocable by Automationdoes not use a trigger and expects either a Flow or a Scheduled Full Calculation to determine when to recalculate.
We often see people doing a mix. They put the Rollup in Scheduled mode but use the Scheduled Full Calculate, seeing the obvious mirror in the word Scheduled. This is going to create the situation I think you're seeing. DLRS is creating the Scheduled Item records but nothing is consuming them. Alternatively, you're running full calculates so you're getting the correct rollup values.
This situation isn't blocked because there are very good reasons to do it but if you want to only use full calculate then switch to Process Builder/Invocable by Automation instead. Alternatively, you may see better time to accuracy and lower system load by leaning into the Invocable by Automation and adding the RollupJob Apex class. (Starting in v2.21 the new LWC wizard has a GUI to help schedule the RollupJob to process those records)
