Fix flaky TestSendCustomMessagePendingError

[tpaschalis]

I'm not entirely sure this is the correct way to test this, so any feedback is appreciated cc @andykellr

The flake is easy to reproduce currently with go test -run=TestSendCustomMessagePendingError -count=1000 -race .

After this fix, I've given it >25k runs and didn't manage to get any failures at all.

Background

The flakiness in TestSendCustomMessagePendingError was due to the fact that the second call to SendCustomMessage was often not fast enough to trigger the condition, giving the first call to c.sender.ScheduleSend() enough time for a message to be sent and clearing the previous CustomMessage from the sender's first queued message in the process.

Solution

Without going deeper to add explicit synchronization, I think that delaying the start to the server can guarantee that no messages are sent out at the beginning, and starting it up later is stil valid for the rest of the assertions

Fixes #278

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 76.32%. Comparing base (ed38d5f) to head (6bba093).
Report is 4 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #292   +/-   ##
=======================================
  Coverage   76.32%   76.32%           
=======================================
  Files          25       25           
  Lines        1681     1681           
=======================================
  Hits         1283     1283           
  Misses        291      291           
  Partials      107      107           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[tpaschalis]

Any feedback here?

[tpaschalis]

cc @andykellr hey there Andy👋 sorry for the ping, but do you have any feedback on this?

[andykellr]

Looks great, thank you for the fix!