#4812 adds a new table for temporarily storing unprocessed headers in full-sync. This table could be merged with the main headers table, to avoid having to move headers to the main table during full-sync.

I think this will simplify things and remove the (wrong since snap-sync was implemented) invariant of "we have a header in DB -> we have its block in DB". An example of this invariant at play is the gas estimator:

2025-10-17T09:09:50.010242Z ERROR ethrex_rpc::eth::gas_tip_estimator: Block body for block number 23594386 is missing but is below the latest known block!