Investigate unbounded growth issues in net request

[ryardley]

⚠️ Potential issue | 🟠 Major

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
# Inspect OnceTake API to confirm the best way to consume the channel.
rg -n "struct OnceTake|impl OnceTake|OnceTake<" -S --type=rust

Repository: gnosisguild/enclave

Length of output: 656

---

🏁 Script executed:

cat -n crates/utils/src/helpers.rs | sed -n '40,100p'

Repository: gnosisguild/enclave

Length of output: 916

---

🏁 Script executed:

cat -n crates/net/src/net_sync_manager.rs | sed -n '148,175p'

Repository: gnosisguild/enclave

Length of output: 1166

---

🏁 Script executed:

rg -n "self\.requests" crates/net/src/net_sync_manager.rs

Repository: gnosisguild/enclave

Length of output: 189

---

🏁 Script executed:

rg -n "requests\[|requests\.get|requests\.remove|requests\.insert" crates/net/src/net_sync_manager.rs

Repository: gnosisguild/enclave

Length of output: 189

---

🏁 Script executed:

cat -n crates/net/src/net_sync_manager.rs | sed -n '1,150p' | tail -50

Repository: gnosisguild/enclave

Length of output: 2090

---

🏁 Script executed:

rg -n "Handler.*SyncRequestReceived|Handler.*ReceiveEvents" crates/net/src/net_sync_manager.rs -A 2

Repository: gnosisguild/enclave

Length of output: 417

---

🏁 Script executed:

rg -n "impl Drop|impl.*Cleanup|requests\.clear\|fn drop" crates/net/src/net_sync_manager.rs

Repository: gnosisguild/enclave

Length of output: 45

---

🏁 Script executed:

rg -n "ReceiveEvents|struct ReceiveEvents" crates/net/src/ -B 2 -A 5

Repository: gnosisguild/enclave

Length of output: 1564

---

🏁 Script executed:

rg -n "struct ReceiveEvents" crates/net/src/ -A 10

Repository: gnosisguild/enclave

Length of output: 45

---

🏁 Script executed:

rg -n "ReceiveEvents" crates/ -A 5 | grep -A 5 "struct ReceiveEvents"

Repository: gnosisguild/enclave

Length of output: 364

---

🏁 Script executed:

cat -n crates/net/src/net_sync_manager.rs | sed -n '148,172p'

Repository: gnosisguild/enclave

Length of output: 1079

---

Remove fulfilled requests from requests to avoid unbounded growth.

Entries are never cleared after a response is sent, causing long-running nodes to accumulate stale entries indefinitely. Change .get() to .remove() to delete the entry from the HashMap once the response is sent. Since channel.to_owned() clones the OnceTake before removal, the sent channel remains functional.

Suggested change

-            let Some(channel) = self.requests.get(&msg.id()) else {
+            let Some(channel) = self.requests.remove(&msg.id()) else {
                 bail!("request not found with {}", msg.id());
             };

🤖 Prompt for AI Agents

In `@crates/net/src/net_sync_manager.rs` around lines 148 - 167, The handler for
ReceiveEvents in NetSyncManager currently uses self.requests.get(&msg.id()),
leaving entries in the requests map and causing unbounded growth; change the
lookup to remove the entry (use self.requests.remove(&msg.id())) so the request
is deleted once handled, but still clone the channel for sending
(channel.to_owned()) before/after removal as needed to preserve the OnceTake
value, and then proceed to send the NetCommand::SyncResponse via
self.tx.try_send as before.

Originally posted by @coderabbitai[bot] in #1153 (comment)