Function cached_parse 

pub fn cached_parse(
    path: &Path,
    content: &str,
) -> (FileId, Result<(Arc<Vec<SourceUnit>>, Arc<Vec<CompileError>>), Arc<Vec<CompileError>>>)

The strict parse of path’s content — parser::parse_units_with_warnings_from, cached by content equality and keyed on the durable FileId file_id_for assigns path. On a cache hit, no new ExprIds are consumed (the cached units already carry their own, fixed at whenever they were last actually parsed); on a miss, [DECISION D]’s durable counter advances by however many the fresh parse used.

The lock is held across the parse itself (not released and re-acquired the way PROJECT_UNIT_CACHE did around its own parse) — [DECISION D]’s counter must be threaded through the parse call, and releasing the lock mid-parse would need a pre-reserved ExprId block of unknown size. Serialises concurrent parses process-wide; each individual file’s parse is fast enough (sub-millisecond, ordinarily) that this is a deliberate, documented trade of a little concurrency for not having to invent a block-reservation scheme — revisit only if profiling ever shows real contention. See file_id_for’s own doc comment for why the lock’s unwrap_or_else(PoisonError::into_inner) here is safe.