test(harness): widen internal request deadline for shared CI runners

A 3-node cluster packed onto ~2 shared CI cores runs several times
slower than real hardware, so the production 30s internal request
deadline could spuriously abort cross-shard transactions that
complete in ~10s locally. Raise it for harness-spawned nodes only,
staying below nextest's cluster terminate ceiling so a genuine hang
still fails cleanly.

Author: farhan-syah

nodedb-test-support/src/cluster_harness/node/lifecycle/spawn_full.rs

@@ -24,6 +24,15 @@ use crate::cluster_harness::cluster::ClusterSpawnConfig;
 use super::client_slot::ClusterTestClient;
 use super::types::{DataDir, TestClusterNode};
 
+/// Internal request deadline for harness-spawned nodes, widened well above the
+/// production 30s default. Cross-shard work on a CI runner that packs a 3-node
+/// cluster onto ~2 shared cores runs several times slower than on real
+/// hardware, so the production deadline spuriously aborts transactions that
+/// complete in ~10s locally. This value sits above the slowest observed
+/// cross-node operation yet below nextest's cluster terminate ceiling, so a
+/// genuine hang still fails cleanly instead of being masked.
+const TEST_REQUEST_DEADLINE_SECS: u64 = 120;
+
 impl TestClusterNode {
     /// [`Self::spawn_with_full_config`] entry point used by every spawn
     /// variant that does not need to reopen an existing data directory: mints
@@ -165,6 +174,21 @@ impl TestClusterNode {
         // Wire cluster handles into SharedState (mirrors main.rs).
         // `Arc::get_mut` is valid here: `shared` has not been cloned.
         if let Some(state) = Arc::get_mut(&mut shared) {
+            // Widen the internal request deadline for the test environment. In
+            // production a node owns real cores; here a 3-node cluster is packed
+            // onto a CI runner with as few as 2 shared cores (one Data-Plane
+            // core per node, plus each node's Tokio/Raft), so a heavy cross-shard
+            // transaction — e.g. an OLLP predicate delete that resolves a
+            // surrogate per matched edge across nodes — runs several times slower
+            // than on real hardware and can exceed the production 30s deadline
+            // even though it completes in ~10s locally. This is the same class of
+            // timing compensation the harness already applies via fast-election
+            // `ClusterTransportTuning`: the internal request deadline is the one
+            // knob that was still left at its production default. It only ever
+            // GRANTS headroom, so it cannot hide a regression — a genuine hang
+            // is still terminated by nextest's cluster ceiling (see
+            // `.config/nextest.toml`), which is lower than this deadline.
+            state.tuning.network.default_deadline_secs = TEST_REQUEST_DEADLINE_SECS;
             state.node_id = handle.node_id;
             state.cluster_topology = Some(Arc::clone(&handle.topology));
             state.cluster_routing = Some(Arc::clone(&handle.routing));