Build each side's digest-to-Evidence index once instead of scanning native observations for every comparison item. Retain every exact match and avoid linking subjectless Evidence to nodes lacking content identity.
Comparison and relationship regressions pass. Isolated alternating unprofiled runs at 2,000 nodes and observations per side have conventional medians 379.924 ms before and 246.466 ms after; the complete 4,000-item outputs match byte-for-byte. Timings are bounded workload evidence, not whole-suite performance.
Flameox profile: flameox://evidence/7dca714833dc3d254a75704670a81de656b4b31d4b77a72236eecb82459b3d86
compareJavaScriptApplicationVersions built the change graph with
createJavaScriptApplicationGraph, which validates it and re-derives
every node, observation, edge and graph ID, then parsed the result
with a schema that embeds the full graph schema, repeating all of
that work. Seal the change graph with the owned immutable factory and
accept that exact sealed graph in the result parse, as owned analysis
results already do. Items, coverage and the public result schema are
unchanged; output is byte-identical.
morluto