diff --git a/packages/audiodocs/docs/utils/file-concatenation.mdx b/packages/audiodocs/docs/utils/file-concatenation.mdx
index 36c03a7bf..41b93f802 100644
--- a/packages/audiodocs/docs/utils/file-concatenation.mdx
+++ b/packages/audiodocs/docs/utils/file-concatenation.mdx
@@ -21,12 +21,17 @@ recording is finished.
 :::
 
 :::caution
-**Only WAV, M4A, and FLAC concatenation is supported** — the recording formats
+**Only WAV, M4A, FLAC, and ADTS concatenation is supported** — the recording formats
 available on both iOS and Android.
 
 WAV and FLAC decode the segments and write them out again through the system
 encoder — WAV as 32-bit float PCM, FLAC losslessly. M4A remuxes AAC packets
 with system APIs — iOS AVFoundation and Android MediaExtractor/MediaMuxer.
+
+ADTS (`.aac`) appends complete encoded frames without decoding or re-encoding.
+Inputs must use the same sample rate and channel configuration. An incomplete
+trailing frame, such as one left by an app crash, is discarded before the next
+file's frames are appended.
 :::
 
 :::tip Segment size and quality
@@ -68,7 +73,8 @@ come from the same recorder configuration:
 
 Use an output path with an extension that matches the input files, for example
 joining `.m4a` segments into an `.m4a` output file, `.wav` segments into a
-`.wav` output file and so on.
+`.wav` output file, `.flac` segments into a `.flac` output file, or ADTS `.aac`
+segments into an `.aac` output file.
 
 <details>
 <summary>Example usage with rotated recordings</summary>
diff --git a/packages/react-native-audio-api/android/src/main/cpp/audioapi/android/AndroidEncoder.cpp b/packages/react-native-audio-api/android/src/main/cpp/audioapi/android/AndroidEncoder.cpp
index 1f7abe592..9b9176ffc 100644
--- a/packages/react-native-audio-api/android/src/main/cpp/audioapi/android/AndroidEncoder.cpp
+++ b/packages/react-native-audio-api/android/src/main/cpp/audioapi/android/AndroidEncoder.cpp
@@ -686,7 +686,7 @@ class AdtsBackend : public MediaCodecBackend {
     }
     samplingFrequencyIndex_ = *adts::samplingFrequencyIndex(static_cast<int>(desired.sampleRate));
 
-    file_ = std::fopen(filePath.c_str(), "wb");
+    file_.reset(std::fopen(filePath.c_str(), "wb"));
     if (file_ == nullptr) {
       return "AdtsBackend: failed to open file for writing";
     }
@@ -698,8 +698,7 @@ class AdtsBackend : public MediaCodecBackend {
         static_cast<int>(encoding.bitRate),
         true);
     if (!err.empty()) {
-      std::fclose(file_);
-      file_ = nullptr;
+      file_.reset();
       return err;
     }
     return "";
@@ -731,28 +730,30 @@ class AdtsBackend : public MediaCodecBackend {
       return "AdtsBackend: encoded frame too long for an ADTS header";
     }
 
-    if (std::fwrite(header->data(), 1, header->size(), file_) != header->size() ||
-        std::fwrite(data + info.offset, 1, payloadSize, file_) != payloadSize) {
+    if (std::fwrite(header->data(), 1, header->size(), file_.get()) != header->size() ||
+        std::fwrite(data + info.offset, 1, payloadSize, file_.get()) != payloadSize) {
       return "AdtsBackend: short write";
     }
     // Flushed per frame, so the crash-recoverability the format exists for holds at frame
     // granularity rather than at stdio's buffer size.
-    std::fflush(file_);
+    if (std::fflush(file_.get()) != 0) {
+      return "AdtsBackend: failed to flush file";
+    }
     bytesWritten_ += header->size() + payloadSize;
     return "";
   }
 
   std::string finalize() override {
     codec_.reset();
-    if (file_ != nullptr) {
-      std::fclose(file_);
-      file_ = nullptr;
+    if (file_ != nullptr && std::fclose(file_.release()) != 0) {
+      return "AdtsBackend: failed to close file";
     }
     return "";
   }
 
  private:
-  std::FILE *file_{nullptr};
+  // Also closes the file when codec draining fails before finalize() is reached.
+  std::unique_ptr<std::FILE, decltype(&std::fclose)> file_{nullptr, &std::fclose};
   uint8_t samplingFrequencyIndex_{0};
   size_t bytesWritten_{0};
 };