Keep the read-and-reply toll test inside the network reward window #85

Problem

In Actions run 37319184103, should receive toll on read and on reply failed all three attempts. The replies were accepted, but were sent after the network's 60-second reward window, so the test incorrectly expected the reply reward.

The initial-message-to-reply transaction timestamp differences were 60,333 ms, 60,502 ms, and 60,837 ms. Expected final balance: 52.437500; actual: 49.343750. The missing 3.093750 LIB matches the taxed half-toll reward. The server's reward deadline starts at the original message timestamp, not when it is read.

Location and cause

playwright-tests/tests/multiUser.e2e.test.js, test should receive toll on read and on reply:

await sendMessageTo(pg2, user1, msg2);
await pg1.waitForTimeout(10_000);
await checkReceivedMessage(pg1, user2, msg2);
// wallet navigation...
await pg1.waitForTimeout(20_000);
// check read reward, then reply
await sendMessageTo(pg1, user2, replyMsg);

sendMessageTo() also waits ten seconds for chat initialization before sending. Together, fixed pauses, navigation, and transaction processing consume the reward window. This is a test timing problem; the observed server behavior is consistent with the configured deadline.

Resolution

• Wait for the received message through the existing bounded UI assertion instead of sleeping before checking it.
• Poll the expected read-reward balance using the existing wallet helper, then reply immediately after it is confirmed.
• Poll the final balance to verify the reply reward instead of an unconditional post-reply sleep.
• Preserve the full-precision toll setup and separate exact six-decimal assertions for both rewards.
• Keep shared chat readiness safeguards and unrelated tests unchanged.

Acceptance

Run this test repeatedly against https://liberdus.com/dev/ with retries disabled. Both reward assertions must pass; inspect transaction timing to ensure the reply is comfortably inside the reward window. Stack the fix above test/settle-attachment-menu-scroll, then rebase test/retain-failure-diagnostics onto the fix.

This is the immediate fix: unlike the other three intermittent findings from this run, it failed every attempt.