fix(stage_baches): rpc resequence stop stage on unwind

[V-Staykov]

No description provided.

[sonarcloud]

Quality Gate passed

Issues
2 New issues
0 Accepted issues

Measures
0 Security Hotspots
22.2% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarCloud

[V-Staykov]

@hexoscott I removed an unneeded sleep and now only one is left that is waiting 1s between checks for new block on the datastream. I ran locally a sequencer + rpc and it doesn't skip blocks for me.

[hexoscott]

Awesome! Working perfectly now with no delay between sequencer and RPC

[cffls]

Resequence test is passing now. @V-Staykov please take a look at the fix. I am not sure if this is the optimal solution, because the RPC node took longer to catch up the reorg so I had to increase the timeout.

Here is how I came up with the fix:

• Download the RPC log from previous failed run: https://github.com/0xPolygonHermez/cdk-erigon/actions/runs/11708016946
• The node detected that DS was behind, which meant a reorg has happened.
• Then the node immediate failed with an error when trying to find the latest common ancestor. [cdk-erigon-node-001] [EROR] [11-06|18:50:07.103] Error while executing stage err="[3/15 Batches] GetL2BlockByNumber: failed to get l2 block 50 from datastream: getL2BlockByNumber: initiateDownloadBookmark: afterStartCommand: readResultEntry: client already started: Already started".

From the error, it seems clear that the DS connection established before should be stopped before setting a new bookmark.