Recorded queries include the startTime from the customer cluster.

The replay client doesn't currently respect this; it will issue commands as quickly as possible.

The client should identify the start time of the entire recording in the customer cluster, and note the start time of the replay, and should replay queries at the correct offset.

This could be achieved by

• Inspecting the start time of the first query - potentially requires scanning multiple chunks of queries, as multiple threads will record queries in local buffers concurrently but only persist when the buffer fills so the "first" query may be in a chunk which was not written until later
• Adding a new record type which persists metadata relevant to the whole recording, like the absolute start time
• Changing query start/end times to be relative to the start time of the recording, managed by the customer cluster.