To provide integrity checks, the recording should compute a checksum of all data in a given recording file.

Ideally the checksum would be computed incrementally from the data to be recorded for each "event"/"packet", rather than as a post-completion step involving reading the file back from disk. However, this can be benchmarked.

S3 provides object integrity checks; we should select a supported checksum function. We do not need a cryptographically secure hash; we do not need to guard against malicious action, just reduce the likelihood of a partial or corrupted recording file causing unexpected failures during replay.

If the checksum is to be used for S3 integrity checks, it will need to be stored outside of the file to be uploaded to S3.

As a recording will be composed of several files, the checksums may either be stored along side the recording files (e.g., "foobar-001.recording" and "foobar-001.checksum") or in an index listing checksums for all files in a recording.

Traffic recordings write to the output in traffic_recorder.cpp. At this point, the data which will be written is available in memory.