This PR fixes errorCodes for Hive tests:
- eth_getLogs/filter-error-future-block-range (erigon_default)
- eth_getLogs/filter-error-reversed-block-range (erigon_default)

And align few errors code with Geth

waiting workflow on Hive tests is completed

…truct (#18269) (#18314)

This PR refactors the blob history downloading logic from
`downloadBlobHistoryWorker` function into a dedicated
`BlobHistoryDownloader` struct in the `cl/phase1/network` package. This
provides better encapsulation, testability, and allows for dynamic
control of the blob backfilling process.

## Changes

### New: `cl/phase1/network/blob_downloader.go`

Introduces `BlobHistoryDownloader` struct with the following features:

- **Configurable head slot**: The `headSlot` can be updated via
`SetHeadSlot()` to target higher slots as the chain progresses
- **Periodic downloading**: Runs a download loop every 12 seconds when
started
- **Peer count check**: Skips iterations if peer count is below 16 and
logs a warning
- **Conditional start**: Only starts the goroutine if `archiveBlobs` or
`immediateBlobsBackfilling` is enabled
- **Progress tracking**: Maintains `highestBackfilledSlot` to track
backfilling progress
- **Callback support**: `SetNotifyBlobBackfilled()` allows setting a
callback when backfilling completes

#### Key Methods

| Method | Description |
|--------|-------------|
| `NewBlobHistoryDownloader()` | Constructor with all dependencies | |
`SetHeadSlot(slot)` | Sets the target head slot (currentSlot + 1) | |
`SetNotifyBlobBackfilled(fn)` | Sets completion callback | |
`HeadSlot()` | Returns current head slot |
| `HighestBackfilledSlot()` | Returns highest backfilled slot | |
`Running()` | Returns whether downloader is active | | `Start()` |
Begins the download loop |

### Modified: `cl/phase1/stages/clstages.go`

- Added `blobDownloader` field to `Cfg` struct
- Updated `ClStagesCfg()` to accept `ctx context.Context` parameter
- Creates `BlobHistoryDownloader` instance in the config constructor

### Modified: `cl/phase1/stages/stage_history_download.go`

- Added `blobDownloader` field to `StageHistoryReconstructionCfg`
- Updated `StageHistoryReconstruction()` to accept `blobDownloader`
parameter
- Removed inline `downloadBlobHistoryWorker` function (~140 lines)
- Stage now uses the passed-in `blobDownloader` instance

## Migration Notes

This is a refactor with no behavioral changes to the blob downloading
logic. The same functionality is preserved but now encapsulated in a
reusable struct. Additionally, we can defer blob download everwhere
except for gossip

issue #18308

https://beacon.chiadochain.net/slot/20031440
target epoch is `1251964` , but in our codebase, committee count is
computed by current epoch `1251965`

…8315)

See #18302

(cherry picked from commit b202139)

---------

Co-authored-by: milen <[email protected]>
Co-authored-by: Copilot <[email protected]>

Previous limit was not enough for a 6 blob EIP-4844 transaction with
commitments and proofs. The theoretical max size is ~824 KB.

… peer drops (#18264)

for #18217 and
#18263

it's enough if they remain within min supported and current version.

cherry-pick of #18374

---

when exec responds with !ready (say due to pruning) Caplin should try
again before it gives up

that seems to be the intent in the code but the implementation has a bug
- it uses a `Timer` instead of a `Ticker`

we're getting missed (orphaned) blocks due to this

This fixes: #18436

Post the gnosis balancer HF block initialization requires a parent read
which fails for blocks in snapshots without this change

- #18397 was reverted because it broke sync
- root cause: the receipts were produced fine, but txnum-2 execution was
writing accounts in domains, leading to state root mismatch
- in this PR, the re-execution of txnum-1 is removed; we can generate
correct receipts simply from Receipt domain. No re-exec needed.

Enables discv5 for 3.3 for
#18195. Duplicates #18578
from main.

I think we want discv5 enabled in 3.3 if we want to meet the Jan 31
target. I wonder if we want to keep the resolution leaning toward discv4
for 3.3.