Investigate relatively high p95 aggregated storage delete latency on GitLab.com

Context

Followup from 2024-04-05: RegistryServiceDatabaseApdexSLOViol... (gitlab-com/gl-infra/production#17785 - closed). See this thread for more details.

Task

We have observed that the p95 for the aggregated storage delete metric is hovering around 1s, while the online GC blob delete metrics are hovering around 500ms. Given that most deletes now occur on the metadata DB and not the storage backend, we should investigate why the former is significantly higher than the latter.