Quick answer: Unreal cooks suddenly take an order of magnitude longer? Shared Derived Data Cache (DDC) might be returning corrupted entries that fail validation, forcing re-derivation.

A team's CI cook went from 20 minutes to 4 hours after the shared DDC server hiccupped. Each asset hits the DDC, gets bad data, and re-cooks from source.

DDC Hit / Miss Stats

Run cook with -LogCmds="LogDerivedDataCache Verbose". Inspect hit / miss / fail rates. High fails indicate corruption; high misses indicate cache scope.

Verify and Purge

Engine ships UnrealEditor -DDCDumpSizeFor=Shared to inspect and validate. Suspect entries can be purged; legitimate entries are re-cached on next access.

Network Filesystem Quirks

SMB / NFS shares can serve partial writes during high contention. Tune your DDC backend; consider switching to S3 / GCS for atomic uploads.

Local DDC Fallback

Each developer has a local DDC layer. If shared is suspect, set BootProcessor=Local in DerivedDataCache.ini temporarily — cooks slower the first time, then fast and clean.

Verifying

Cooks return to baseline duration. DDC hit rate stabilizes at the expected level.

“A corrupt DDC re-derives everything. Validate, purge, or fall back to local.”

Monitor cook time in CI and alert on regressions of more than 1.5x — DDC corruption is much cheaper to catch the day it starts.