A poison message refers to a clientâÂÂserver model issue, where a client machine tries to send a message to the server and fails too many times (the actual amount of "too many" is variable).
The behavior toward poison messages varies - they are either discarded, create a service request event, or initiate other failure indications. The term is used mainly in Microsoft-related frameworks, like SQL Server or Windows Communication Foundation (WCF). RabbitMQ also has a notion of poisoned messages.