Bug triage results: 2026-06-29

[andygrove]

Triage pass over the open requires-triage queue, per the project Bug Triage Guide.

• Total issues processed: 11 (7 triaged, 4 skipped, 0 failed)
• Priority counts applied: priority:critical 2, priority:medium 5
• Guide: docs/source/contributor-guide/bug_triage.md

Labels have already been applied and requires-triage removed from each issue listed under "Triaged". A reviewer should spot-check the calls and close this issue when satisfied. To correct a label, edit the affected issue directly.

Triaged

priority:critical

• Comet produces bloated results in comparison with Spark (#4723)
  • Area labels: area:scan, area:aggregation
  • Rationale: priority:critical was preserved as set by the reporter; SUM aggregation over native Iceberg scans silently returns wrong (inflated) values and the user can compare per-row Spark vs Comet output, matching the guide's decision-tree step 1 (silent wrong results).
• [Bug] percentile: DataFusion quantizes interpolation weight to 6 decimal places (#4719)
  • Area labels: area:expressions, area:aggregation
  • Rationale: silent wrong result vs Spark for the native percentile / median / percentile_cont path (bounded by (upper - lower) * 1e-6, no fallback or warning); decision-tree step 1.

priority:medium

• Comet Native scan in Azure fails with workload identity (ignores ABFS configs and env vars) (#4747)
  • Area labels: area:scan
  • Rationale: native scan drops fs.azure.* configs and AZURE_* env vars so abfss reads fail with HTTP 400 from IMDS; the user can work around it by disabling Comet's native scan, so this is a functional gap with a workaround (decision-tree step 3).
• [Bug] AVG(decimal) over a window always falls back to Spark on Spark 4.x (AvgDecimal window branch is dead) (#4731)
  • Area labels: area:expressions
  • Rationale: results stay correct via Spark fallback; the reporter notes this is a coverage gap, not a correctness divergence, so priority:medium per decision-tree step 3.
• bug: shutdown jni_api::TOKIO_RUNTIME on exit (#4725)
  • Area labels: area:ffi
  • Rationale: occasional JVM hang at shutdown because tokio runtime workers attached to the JVM remain as non-daemon threads; affects benchmark/test exit, not query results, with a kill-the-JVM workaround — functional bug with workaround per decision-tree step 3.
• Support fully-native multi-stage (distinct-combined) collect_list / collect_set (#4724)
  • Area labels: area:expressions, area:aggregation
  • Rationale: multi-stage collect_list/collect_set with a PartialMerge would crash natively (Cast error: Cannot cast LIST to non-list data type Binary), but PR feat: native collect_list / array_agg aggregate #4720 already added fallback guards that route this shape through Spark — so today this is a feature gap with a working fallback, matching decision-tree step 3.
• Table Provider API (#4706)
  • Area labels: area:scan
  • Rationale: enhancement to abstract scan APIs so additional table providers (Delta, etc.) can be supported; missing-feature gap with the existing per-provider integration as a workaround.
• Comet aggregation task crashes with offset overflow Comet aggregation task crashes with offset overflow #4718

Skipped — needs more info

• Bug triage results: 2026-06-22 (#4705)
  • Prior triage summary issue (auto-labeled requires-triage); meta, awaiting human review and closure, not a bug.
• Bug triage results: 2026-06-11 (#4625)
  • Prior triage summary issue (auto-labeled requires-triage); meta, awaiting human review and closure, not a bug.
• Bug triage results: 2026-06-01 (#4548)
  • Prior triage summary issue (auto-labeled requires-triage); meta, awaiting human review and closure, not a bug.
• Bug triage results: 2026-05-26 (#4441)
  • Prior triage summary issue (auto-labeled requires-triage); meta, awaiting human review and closure, not a bug.