Unable to run wt tool to verify the palite-backed database on this patch task: Spruce logs. Patch-linked run appears affected.

Summary

Palite materialization filter uses steady_clock; DBs unreadable after reboot

Description

After a host reboot, opening a palite-backed disagg database can fail even though page-log data is intact. plh_get returns 0 entries for pages that exist in SQLite, and open/verify aborts.

Root cause

GET_PAGES filters with timestamp_materialized_us <= now_us(), and now_us() uses steady_clock (boot-relative). Timestamps from the previous uptime are larger than the post-reboot clock, so every page looks unmaterialized.

Example: stored ~1.49e10 vs now_us() ~1.3e9 → 0 rows, despite the page being present in pages_*.db.

Existing knobs do not help:

• disaggregated.last_materialized_lsn — WT frontier only
• materialization_delay_ms — writes only
• palite last_materialized_lsn — unused on get

Impact

wt verify, wt list, and format -v -R cannot reopen intact palite RUNDIRs after reboot.

Acceptance

Post-reboot open/verify works on an intact palite DB without mutating page-log files. 

Workaround

None via config. Manually zeroing timestamp_materialized_us in pages_*.db restores reads but mutates the DB.