Problem

The final flush at shutdown could skip and permanently lose a file that was being written concurrently.

FilesOrchestrator hides files currently being appended to (activeWrites) from the reader, so the periodic upload worker never reads a half-written file. That's correct mid-session — the file is picked up on the next worker cycle. But at shutdown there is no next cycle: the worker is stopped right after the flush, so any file still in activeWrites during the final enumeration is stranded on disk and never uploaded.

The race window is a write submitted concurrently with the flush (e.g. a span ending on another thread during teardown). Writes submitted before the flush were already safe — the serial writer queue plus the barrier in closeCurrentFile() guarantees they complete and release their activeWrites entry first.

The previous shutdown order made this worse: Exporter.shutdown() flushed first, then stopped the worker — so a skipped file had no chance of being rescued.

Fix

• FileWriter.stop() — permanently seals the writer: drains in-flight writes via the serial-queue barrier, finalizes and closes the current file, then rejects all subsequent writes. Writes after stop are logged with their encoded payload so dropped data is debuggable.
• FeatureStoreAndUpload.stop() — ordered teardown: (1) stop scheduling background uploads, (2) seal the writer, (3) run the final exhaustive flush. After sealing, no file can be in activeWrites, so the flush enumerates a complete, quiescent set — nothing is skipped. Stopping uploads first also keeps the periodic worker from racing the seal.
• Exporter.shutdown() — dropped the redundant pre-shutdown flush; each storage's stop() now does the complete, race-free upload.
• DDTracer.shutdown() — no longer pre-flushes. The OTel processor shutdown drains pending spans into the exporter, and the providers flush spans/logs/telemetry properly on their own. Only the cross-process RUM port flush is forwarded (extracted into flushRUM()), since provider shutdown doesn't cover it.

Verification

• Confirmed the OTel shutdown path drains before tearing down for both SimpleSpanProcessor (serial processorQueue.sync barrier) and BatchSpanProcessor (worker.shutdown() → forceFlush drains the buffer synchronously before spanExporter.shutdown()). The drained writes are enqueued on the writer queue before writer.stop()'s barrier, so they land on disk and get uploaded.
• SimpleLogRecordProcessor is fully synchronous (no buffer).
• Builds; FileWriterTests, EventsExporterTests, SpansExporterTests, LogsExporterTests, DataUploadWorkerTests pass.

🤖 Generated with Claude Code