diff --git a/.github/workflows/tpcds.yml b/.github/workflows/tpcds.yml
index 92b908745..569ce1dfb 100644
--- a/.github/workflows/tpcds.yml
+++ b/.github/workflows/tpcds.yml
@@ -52,10 +52,34 @@ on:
 
 jobs:
   tpcds-sf1:
-    name: TPC-DS SF1 (all queries, static planner)
+    name: TPC-DS SF1 (${{ matrix.label }})
     runs-on: ubuntu-latest
     container:
       image: amd64/rust
+    # A passing leg takes ~7 minutes. Cap the job well under the 6-hour
+    # default so a hung query fails fast and frees the runner rather than
+    # sitting for hours with no useful signal.
+    timeout-minutes: 45
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - label: "AQE off, 16 partitions"
+            planner_args: ""
+            partitions: 16
+            slug: "aqe-off-p16"
+          - label: "AQE off, 1 partition"
+            planner_args: ""
+            partitions: 1
+            slug: "aqe-off-p1"
+          - label: "AQE on, 16 partitions"
+            planner_args: "-c ballista.planner.adaptive.enabled=true"
+            partitions: 16
+            slug: "aqe-on-p16"
+          - label: "AQE on, 1 partition"
+            planner_args: "-c ballista.planner.adaptive.enabled=true"
+            partitions: 1
+            slug: "aqe-on-p1"
     steps:
       - name: Install dependencies
         run: |
@@ -72,6 +96,10 @@ jobs:
           rust-version: stable
 
       - uses: swatinem/rust-cache@c19371144df3bb44fab255c43d04cbc2ab54d1c4 #2.9.1
+        with:
+          # Share the build cache across all matrix legs — the compiled
+          # binaries are identical; only the per-suite CLI args differ.
+          shared-key: tpcds-sf1
 
       - name: Build Ballista binaries
         run: |
@@ -93,6 +121,10 @@ jobs:
             --output-dir "$RUNNER_TEMP/tpcds-data"
 
       - name: Run TPC-DS queries against Ballista cluster
+        # Primary timeout, on the step rather than the job, so that a hang
+        # still runs the log-upload step below and leaves something to
+        # diagnose. The job-level cap is only a backstop.
+        timeout-minutes: 30
         env:
           DATA_DIR: ${{ runner.temp }}/tpcds-data
           WORK_DIR: ${{ runner.temp }}/work
@@ -149,34 +181,28 @@ jobs:
           done
           nc -z 127.0.0.1 50051 || { echo "executor did not start"; exit 1; }
 
-          # Run the suite under the default (static) planner. The tpcds binary
-          # internally loops all non-skipped queries and exits non-zero on any
-          # failure, so a single invocation covers the whole suite.
-          #
-          # The adaptive planner (AQE on) is intentionally not run here yet: it
-          # currently fails many TPC-DS queries with an `EmptyExec invalid
-          # partition` assertion (issue #2047). Re-add an `AQE on` invocation
-          # once that is fixed.
-          run_suite() {
-            local label="$1"; shift
-            echo "::group::[$label] TPC-DS suite"
-            ./target/tpch-ci/tpcds \
-              --host 127.0.0.1 --port 50050 \
-              --path "$DATA_DIR" \
-              --partitions 16 \
-              --verify \
-              "$@"
-            echo "::endgroup::"
-          }
-
-          run_suite "static planner" \
-            -c datafusion.optimizer.prefer_hash_join=false
+          # This matrix leg runs one planner/partitioning configuration over the
+          # whole suite. The tpcds binary internally loops all non-skipped
+          # queries and exits non-zero on any failure, so a single invocation
+          # covers the suite. The other legs run in parallel jobs.
+          echo "::group::[${{ matrix.label }}] TPC-DS suite"
+          ./target/tpch-ci/tpcds \
+            --host 127.0.0.1 --port 50050 \
+            --path "$DATA_DIR" \
+            --partitions ${{ matrix.partitions }} \
+            --verify \
+            -c datafusion.optimizer.prefer_hash_join=false \
+            ${{ matrix.planner_args }}
+          echo "::endgroup::"
 
       - name: Upload cluster logs on failure
-        if: failure()
+        # `!success()` rather than `failure()` so a timed-out or cancelled
+        # run still surfaces its scheduler/executor logs — the hang case is
+        # exactly when they are most worth having.
+        if: ${{ !success() }}
         uses: actions/upload-artifact@v7
         with:
-          name: tpcds-sf1-cluster-logs
+          name: tpcds-sf1-cluster-logs-${{ matrix.slug }}
           retention-days: 14
           path: |
             ${{ runner.temp }}/scheduler.log
diff --git a/benchmarks/src/bin/tpcds.rs b/benchmarks/src/bin/tpcds.rs
index 630e6f2b4..a93b74697 100644
--- a/benchmarks/src/bin/tpcds.rs
+++ b/benchmarks/src/bin/tpcds.rs
@@ -58,8 +58,7 @@ const TABLES: &[&str] = &[
     "web_site",
 ];
 
-/// Queries excluded from the correctness gate, with the reason. The gate runs
-/// under the default (static) planner; this list reflects that configuration.
+/// Queries excluded from the correctness gate, with the reason.
 /// Remove an entry as the underlying cause is fixed.
 const SKIP: &[(usize, &str)] = &[
     // Non-deterministic: LIMIT/ORDER BY ties without a total order make the
@@ -69,17 +68,30 @@ const SKIP: &[(usize, &str)] = &[
         71,
         "non-deterministic (ORDER BY ext_price ties; varies run-to-run)",
     ),
-    // tpcgen-cli's TPC-DS schema uses column names that differ from the
-    // DataFusion (branch-54) query text, so these fail at plan time. Not a
-    // Ballista issue (e.g. cr_return_amount_inc_tax vs cr_return_amt_inc_tax).
-    (64, "tpcgen-cli schema column-name mismatch with query text"),
-    (81, "tpcgen-cli schema column-name mismatch with query text"),
-    (84, "tpcgen-cli schema column-name mismatch with query text"),
-    (85, "tpcgen-cli schema column-name mismatch with query text"),
-    (93, "tpcgen-cli schema column-name mismatch with query text"),
-    // Note: the adaptive planner (AQE on) additionally fails many queries with an
-    // `EmptyExec invalid partition` assertion (issue #2047); the gate runs the
-    // static planner, so those are not listed here.
+];
+
+/// Column renames applied to the query text before planning, as
+/// `(query text, tpcgen-cli)` pairs.
+///
+/// `tpcgen-cli` names three columns differently from the TPC-DS spec, and the
+/// DataFusion query text follows the spec, so these queries would otherwise
+/// fail at plan time with "column not found". These are pure renames -- the
+/// data is present under the other name -- so rewriting the reference is
+/// enough. The identical rewrite is applied to the Ballista and oracle runs,
+/// so the comparison stays apples-to-apples.
+///
+/// This is applied at load time rather than by editing the files under
+/// `benchmarks/queries-tpcds/`, because `dev/vendor-tpcds-queries.sh`
+/// re-downloads all 99 queries and would clobber any local edit.
+///
+/// Drop a pair once `tpcgen-cli` renames the column to its spec name.
+const COLUMN_RENAMES: &[(&str, &str)] = &[
+    // income_band: affects q64, q84.
+    ("ib_income_band_sk", "ib_income_band_id"),
+    // reason: affects q85, q93.
+    ("r_reason_desc", "r_reason_description"),
+    // catalog_returns: affects q81.
+    ("cr_return_amt_inc_tax", "cr_return_amount_inc_tax"),
 ];
 
 #[derive(Debug, StructOpt)]
@@ -138,6 +150,38 @@ fn split_statements(contents: &str) -> Vec<String> {
         .collect()
 }
 
+/// Replace whole-word occurrences of `from` with `to`.
+///
+/// A plain `str::replace` is not safe here: `r_reason_desc` is a prefix of
+/// `r_reason_description`, so a substring replace would corrupt an already
+/// correct reference. A character is part of a word if it is alphanumeric or
+/// `_`, which matches how SQL identifiers are spelled in these queries.
+fn replace_word(haystack: &str, from: &str, to: &str) -> String {
+    let is_word = |c: char| c.is_alphanumeric() || c == '_';
+    let mut out = String::with_capacity(haystack.len());
+    let mut rest = haystack;
+    while let Some(pos) = rest.find(from) {
+        let (before, after) = rest.split_at(pos);
+        let tail = &after[from.len()..];
+        let boundary_ok = !before.chars().next_back().is_some_and(is_word)
+            && !tail.chars().next().is_some_and(is_word);
+        out.push_str(before);
+        out.push_str(if boundary_ok { to } else { from });
+        rest = tail;
+    }
+    out.push_str(rest);
+    out
+}
+
+/// Rewrite spec column names to the names `tpcgen-cli` actually emits.
+fn apply_column_renames(sql: &str) -> String {
+    COLUMN_RENAMES
+        .iter()
+        .fold(sql.to_string(), |acc, (from, to)| {
+            replace_word(&acc, from, to)
+        })
+}
+
 fn get_query_sql(query: usize) -> Result<Vec<String>> {
     let possibilities = [
         format!("queries-tpcds/q{query}.sql"),
@@ -146,7 +190,9 @@ fn get_query_sql(query: usize) -> Result<Vec<String>> {
     let mut errors = vec![];
     for filename in &possibilities {
         match fs::read_to_string(filename) {
-            Ok(contents) => return Ok(split_statements(&contents)),
+            Ok(contents) => {
+                return Ok(split_statements(&apply_column_renames(&contents)));
+            }
             Err(e) => errors.push(format!("{filename}: {e}")),
         }
     }
@@ -327,4 +373,53 @@ mod tests {
         let skip: &[(usize, &str)] = &[(1, "should be overridden")];
         assert_eq!(selected_queries(Some(1), skip), vec![1]);
     }
+
+    #[test]
+    fn replace_word_only_matches_whole_identifiers() {
+        assert_eq!(
+            replace_word("a.r_reason_desc, b", "r_reason_desc", "x"),
+            "a.x, b"
+        );
+        // A longer identifier that merely contains the needle is left alone.
+        assert_eq!(
+            replace_word("r_reason_desc_2", "r_reason_desc", "x"),
+            "r_reason_desc_2"
+        );
+        assert_eq!(
+            replace_word("my_r_reason_desc", "r_reason_desc", "x"),
+            "my_r_reason_desc"
+        );
+    }
+
+    #[test]
+    fn column_renames_are_idempotent() {
+        // `r_reason_desc` is a prefix of its replacement, so a second pass must
+        // not extend it again.
+        let once = apply_column_renames("select r_reason_desc from reason");
+        assert_eq!(once, "select r_reason_description from reason");
+        assert_eq!(apply_column_renames(&once), once);
+    }
+
+    #[test]
+    fn column_renames_cover_every_spec_name() {
+        let sql = "ib_income_band_sk, r_reason_desc, cr_return_amt_inc_tax";
+        assert_eq!(
+            apply_column_renames(sql),
+            "ib_income_band_id, r_reason_description, cr_return_amount_inc_tax"
+        );
+    }
+
+    #[test]
+    fn renames_are_applied_when_loading_a_query() {
+        // q93 references `r_reason_desc`; after loading, only the tpcgen-cli
+        // spelling should remain. Guards against the rewrite being dropped
+        // from the load path.
+        let Ok(sqls) = get_query_sql(93) else {
+            // Query files are not present in every build context.
+            return;
+        };
+        let joined = sqls.join(" ");
+        assert!(joined.contains("r_reason_description"));
+        assert!(!replace_word(&joined, "r_reason_desc", "?").contains('?'));
+    }
 }