more robust fork indexing

[qlp]

there are often (multiple times a week) re-organized blocks in the index that should be re-indexed. This happens most of the time with fast-moving chains (like bArtio Testnet).

as a result we expose invalid data

root cause could be that we're ofter crashing and then we'll miss re-orged blocks, so maybe better first change error handling so the entire hubble process does not restart upon a failure