When upgrading the platform to 2026.1, we waited for around two hours in total (SF and IAM on DEV) for the TSF optimizer to finish rebuilding the indexes. This was even after old branches/applications were removed in SF and IAM the day before to speed up the upgrade. At one point, we saw the slow_query_log index being rebuilt, where fragmentation was only 0.1%, containing over 5 million records. Our idea is twofold:
- Parameterize or hardcode the index optimization so that only indexes with a fragmentation percentage above a given threshold are rebuilt.
- Allow configuration in the Deployer to specify whether indexes need to be optimized. We already have a nightly job that performs this, so we would prefer to let that run outside office hours.
