fix(harness): disable SeiDB state store on nightly nodes, drop dead ss write-mode

[bdchatham]

Problem

Nightly load suite hangs: the RPC follower's seid parks all threads on futex_wait_queue right after Found 0 WALs, opens no listeners (/proc/net/tcp has no :8545/:26657/:26656), so the benchmark's EVM-readiness gate never passes and NightlyRunFailed fires.

Root cause (grounded in canonical sei-chain main @ 51eb1fd, the nightly image's commit)

• storage.state_store.write_mode no longer exists — sei-db/config/ss_config.go's StateStoreConfig has no write-mode field (EVM routing is the evm-split bool); seid's app.toml template renders no ss-write-mode. The harness's ss-write-mode=memiavl_only was a silently-ignored no-op.
• The real trigger is ss-enable: DefaultStateStoreConfig{Enable: true} — the latest image defaults the SeiDB state store on for full nodes. Enabling it on a fresh-genesis RPC follower deadlocks seid at store-open. Validators are fine because the controller role-defaults them to ss-enable=false (confirmed live: follower ss-enable=true, validator ss-enable=false; everything else identical).

Fix — align the override with the current config schema

In memiavlStorageConfig (applied to validators + followers):

• drop storage.state_store.write_mode (dead key)
• add storage.state_store.enable: "false" (string→bool coerces via sei-config's typed registry, same as the existing evm.worker_pool_size: "32" int override)
• keep storage.state_commit.write_mode: memiavl_only

This matches the healthy validators and costs no FlatKV-migration coverage — that lives in the SC layer (sc-write-mode), not the historical state store. The major-upgrade suite still omits the map.

Validation

• go vet -tags integration ./test/integration/ passes; gofmt clean.
• Post-merge: new harness image → bump cronjob → re-trigger load → confirm seid opens :8545, the follower's startup log shows the SS store disabled (not wedged), and seiload emits seiload_run_duration_seconds.

A separate sei-chain robustness bug (a fresh-genesis ss-enable=true full node should boot or fail-fast, not silently deadlock) is filed independently.

🤖 Generated with Claude Code

[cursor]

PR Summary

Low Risk
Integration-test provisioning config only; aligns followers with validator ss-enable=false and removes a no-op key.

Overview
Fixes nightly integration hangs where RPC followers never open EVM/Tendermint listeners because SeiDB state store defaults on for full nodes on the current image.

memiavlStorageConfig (load, release, chaos suites) now sets storage.state_store.enable: "false" instead of the obsolete storage.state_store.write_mode key, which sei-chain no longer reads. storage.state_commit.write_mode: memiavl_only is unchanged. Comments document the deadlock on fresh-genesis followers and that FlatKV migration coverage stays on the SC layer.

Major-upgrade suite still omits this map.

^{Reviewed by Cursor Bugbot for commit 05ccb5f. Bugbot is set up for automated code reviews on this repo. Configure here.}