I believe CI failures are due to two problems in the tcp mailboxes that I describe them below.
Receiving duplicate commit messages causes negative wait group count and a panic consequently. This issue happened exactly after we decreased send timeout value.
If a remote tcp mailbox aborts after sending precommit causes the receiving side to get blocked forever and it cannot gets closed.
I believe CI failures are due to two problems in the tcp mailboxes that I describe them below.
Receiving duplicate commit messages causes negative wait group count and a panic consequently. This issue happened exactly after we decreased send timeout value.
If a remote tcp mailbox aborts after sending precommit causes the receiving side to get blocked forever and it cannot gets closed.