op-node: syncStatus can report `LocalSafeL2` of zero even when `SafeL2` is nonzero

[geoknee]

Since https://github.com/ethereum-optimism/optimism/pull/14265/files, we are seeing batchers on some networks try to load all blocks from genesis from the sequencer. Rolling back to an old batcher version fixes the problem, where a simple restart does not.

t=2025-02-27T13:23:24+0000 lvl=info msg="Loading range of multiple blocks into state" start=1 end=10057590
t=2025-02-27T13:23:24+0000 lvl=warn msg="next safe block is below oldest block in state" syncActions="SyncActions{blocksToPrune: 0, channelsToPrune: 0, clearState: 0x0000000000000000000000000000000000000000000000000000000000000000:0, blocksToLoad: &{1 10057590}}" oldestBlockInState="&{header:0xc00494cc88 uncles:[] transactions:[0xc0049224e0] withdrawals:[] witness:<nil> hash:{_:[] _:{} v:0xc0006f0c00} size:{_:{} _:{} v:0} ReceivedAt:0001-01-01 00:00:00 +0000 UTC ReceivedFrom:<nil>}" safeL2=0x0000000000000000000000000000000000000000000000000000000000000000:0

[geoknee]

I need to check with the interop team, but I wonder if this is the problem ae50c31

[geoknee]

#14584 is a relevant draft PR which may help solve this one.

[geoknee]

Extra context: the LocalSafeL2=0 case seems to be caused by a change in active sequencer, so could be related to startup behaviour in op-node.

[sebastianst]

If this is still the case, it's definitely a bug. It should be an invariant that cross (un)safe >= local (un)safe. cc @protolambda