Build a reproducer for the large-cache failure mode in which the cache update bytes reach the updates trigger (eviction_updates_trigger) and eviction cannot reduce them effectively, so application threads are pulled into eviction and writes stall. This is the update-trigger counterpart of the dirty-trigger mode tracked in WT-18829.

Goal

• Reproduce the main symptoms: update bytes at or above the updates trigger, low eviction effectiveness on those bytes, application threads evicting.
• Identify which condition keeps update bytes from being evicted (what holds them, and why eviction does not free them).
• From the reproducer, identify quick backportable improvements and statistics, and long-term options for a follow-up project.

Reference cases: HELP-88313 (Santander), HELP-87984, HELP-93123, HELP-101424.

Status: ten reproduction rounds so far (under WT-18829) did not reach the update trigger. Reproduction progress will be recorded on this ticket.

Out of scope: the dirty-trigger mode (WT-18829), implementing a long-term redesign.