[GEN-12] `tokio::time::sleep(50ms)` used to "wait for actor propagation" in ~10 `ClientHandle` methods

[intendednull]

Commit: 2f26d91 · Finding: GEN-12

Problem

crates/client/src/lib.rs contains tokio::time::sleep(Duration::from_millis(50)).await at lines 1141, 1187, 1198, 1234, 1710, 1724, 1753, 1766, 1835, 1920 — all outside #[cfg(test)]. They appear in ordinary ClientHandle methods to "wait for actor propagation" after a do_send.

Problems:

• 50ms is too short on slow machines (races the actor), too long on fast machines (wastes latency).
• A known flakiness source for e2e suites.
• The ask(...) pattern is available and synchronously waits for the actor reply.

Fix

Convert each do_send(...) followed by sleep(50 ms) into an ask(...).await that waits for the actor to apply the mutation. Where the mutation is truly fire-and-forget, use a watch channel or notify.

Obvious? Moderate — per-call-site decision needed but the pattern is mechanical.