In WT-16529 we explored a change where we utilise more of the eviction queue if th queue is frequently empty and we are also at the update eviction trigger. 

However, the change required code that modified scrub eviction behaviour which was causing some regressions (BF-41732). The BF was caused by the removal of the line

bytes_updates < (uint64_t)((updates_target + updates_trigger) * bytes_max) / 200)

We need an alternate method to queue more pages in the situation where we hit eviction triggers.