diff --git a/test/src/utils/threading/concurrent_vector_guard.cpp b/test/src/utils/threading/concurrent_vector_guard.cpp
index 69480855..f4449840 100644
--- a/test/src/utils/threading/concurrent_vector_guard.cpp
+++ b/test/src/utils/threading/concurrent_vector_guard.cpp
@@ -30,9 +30,11 @@
 
 #include "src/common.hpp"
 
+#include "genesis/utils/math/random.hpp"
 #include "genesis/utils/threading/concurrent_vector_guard.hpp"
 #include "genesis/utils/threading/thread_pool.hpp"
 
+#include <atomic>
 #include <ctime>
 #include <vector>
 
@@ -89,3 +91,70 @@ TEST( Threading, ConcurrentVectorGuard )
     }
     EXPECT_EQ( 0, cnt_wrong );
 }
+
+TEST( Threading, VectorEntries )
+{
+    // Test siez
+    size_t const num_threads = 10;
+    size_t const num_vecs = 1000;
+    size_t const max_length = 1000;
+
+    auto const seed = ::time(nullptr);
+    permuted_congruential_generator_init( seed );
+    LOG_INFO << "Seed: " << seed;
+
+    // Fill a vector with vectors of different lengths, with values that are all 1.
+    // Then, we can add up those entries to get our expected number of processed elements.
+    size_t exp_num_elem = 0;
+    auto data = std::vector<std::vector<int>>( num_vecs );
+    for( auto& vec : data ) {
+        vec.resize( permuted_congruential_generator( max_length ), 1 );
+        exp_num_elem += vec.size();
+    }
+
+    // Now we spin up some threads and erase elements in parallel from the vector, starting
+    // at different offsets for speed. Each thread starts processing elements, and removes
+    // them from their vec within data, until empty, and then moves to the next one.
+    auto thread_pool = std::make_shared<ThreadPool>( num_threads );
+    auto vector_guard = ConcurrentVectorGuard( num_vecs );
+    std::atomic<size_t> num_elem = 0;
+    for( size_t t = 0; t < num_threads; ++t ) {
+        thread_pool->enqueue_detached([&, t]()
+        {
+            // Get equally distributed starting group indices.
+            size_t const start_group_idx = t * num_vecs / num_threads;
+            size_t cur_group_idx = start_group_idx;
+
+            while( true ) {
+                // Lock the vector that we are currently operating on.
+                auto lock = vector_guard.get_lock_guard(cur_group_idx);
+
+                // Find the next vector that has data.
+                // If the current one does not, we move to the next (in the next iteration).
+                // If we looped around and arrive back where we started, we are done.
+                if( data[cur_group_idx].empty() ) {
+                    ++cur_group_idx;
+                    cur_group_idx %= num_vecs;
+                    if( cur_group_idx == start_group_idx ) {
+                        break;
+                    }
+                    continue;
+                }
+
+                // Remove the last entry from the vector
+                // LOG_DBG << "thread " << t << " popping data[" << cur_group_idx << "] at " << (data[cur_group_idx].size() - 1);
+                data[cur_group_idx].pop_back();
+                ++num_elem;
+            }
+        });
+    }
+    thread_pool->wait_for_all_pending_tasks();
+
+    // Now we expect to have processed exactly the elements that we put in data,
+    // and that we have left the vectors in data empty.
+    EXPECT_EQ( exp_num_elem, num_elem );
+    for( auto const& vec : data ) {
+        EXPECT_TRUE( vec.empty() );
+        EXPECT_EQ( 0, vec.size() );
+    }
+}