A demand caller (not only the prefetcher) is waiting on this read. Decides the background L2 write's priority: under a full write queue a speculative write is evicted before a demand one.
Fetch-gate class of the shared source read. Raised (never lowered) as more urgent callers join, so a prefetch a frame is now waiting on stops queuing behind every other demand read at the speculative share.
Mutable state for callers sharing one cache read.