diff --git a/src/v/cluster/tx_gateway_frontend.cc b/src/v/cluster/tx_gateway_frontend.cc
index 281a85c51cda6..ba69844cb47ea 100644
--- a/src/v/cluster/tx_gateway_frontend.cc
+++ b/src/v/cluster/tx_gateway_frontend.cc
@@ -1147,7 +1147,7 @@ ss::future<cluster::init_tm_tx_reply> tx_gateway_frontend::limit_init_tm_tx(
 
         // similar to double-checked locking pattern
         // it protects concurrent access to oldest_tx
-        if (stm->tx_cache_size() > _max_transactions_per_coordinator()) {
+        while (stm->tx_cache_size() > _max_transactions_per_coordinator()) {
             auto tx_opt = stm->oldest_tx();
             if (!tx_opt) {
                 vlog(
@@ -1160,7 +1160,10 @@ ss::future<cluster::init_tm_tx_reply> tx_gateway_frontend::limit_init_tm_tx(
             auto tx = tx_opt.value();
             vlog(
               txlog.info,
-              "tx cache is at capacity; expiring oldest tx with id:{}",
+              "tx cache size ({}) is beyond capacity ({}); expiring oldest tx "
+              "with id:{}",
+              stm->tx_cache_size(),
+              _max_transactions_per_coordinator(),
               tx.id);
             auto tx_units = co_await stm->lock_tx(tx.id, "init_tm_tx");
 
@@ -1190,7 +1193,7 @@ ss::future<cluster::init_tm_tx_reply> tx_gateway_frontend::limit_init_tm_tx(
             }
             tx_units.return_all();
         }
-
+        vlog(txlog.info, "tx cache size is reduced");
         init_units.return_all();
     }