diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 95c9bb146..514b7d028 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -160,7 +160,7 @@ jobs:
       matrix:
         # Both physical variants of a workload live in the same file, so they
         # are measured back to back.
-        workload: [ls, dd, tar]
+        workload: [dd, tar]
     env:
       CODSPEED_REV: ${{ github.event.pull_request.head.sha || github.sha }}
       CARGO_TARGET_DIR: ${{ github.workspace }}/target
diff --git a/crates/memtrack/benchmarks/ls.yml b/crates/memtrack/benchmarks/ls.yml
deleted file mode 100644
index de789b0b1..000000000
--- a/crates/memtrack/benchmarks/ls.yml
+++ /dev/null
@@ -1,17 +0,0 @@
-$schema: https://raw.githubusercontent.com/CodSpeedHQ/codspeed/refs/heads/main/schemas/codspeed.schema.json
-
-# Each tracked run pays a fixed BPF load + uprobe attach cost, so the default
-# 1s warmup / 3s max budgets fit only a handful of rounds.
-options:
-  warmup-time: "5s"
-  max-time: "60s"
-
-benchmarks:
-  # Read-only, low-allocation baseline. The tracked command string is run
-  # through `bash -c`, so output can be redirected away: otherwise every round
-  # dumps the whole listing into the runner log.
-  - name: "memtrack track ls"
-    exec: env CODSPEED_MEMTRACK_TRACK_PHYSICAL=0 codspeed-memtrack track "ls -la /usr/bin > /dev/null" --output /tmp/codspeed-memtrack-bench
-
-  - name: "memtrack track ls (with physical)"
-    exec: env CODSPEED_MEMTRACK_TRACK_PHYSICAL=1 codspeed-memtrack track "ls -la /usr/bin > /dev/null" --output /tmp/codspeed-memtrack-bench
diff --git a/crates/memtrack/benchmarks/tar.yml b/crates/memtrack/benchmarks/tar.yml
index f259ecf5e..d96a5bdfa 100644
--- a/crates/memtrack/benchmarks/tar.yml
+++ b/crates/memtrack/benchmarks/tar.yml
@@ -3,7 +3,7 @@ $schema: https://raw.githubusercontent.com/CodSpeedHQ/codspeed/refs/heads/main/s
 # Each tracked run pays a fixed BPF load + uprobe attach cost, so the default
 # 1s warmup / 3s max budgets fit only a handful of rounds.
 options:
-  warmup-time: "5s"
+  warmup-time: "30s"
   max-time: "60s"
 
 # The archive is pre-created by CI before measurement: a first-time multi-GB