When scaling down pods it is possible for transactions to remain in their prepared state. There is an automated procedure for recovering these (see
CLOUD-2210) but there should also be a manual procedure for those cases where automatic recovery fails. The manual procedure for database resources is covered in CLOUD-2238 and a similar document is required for A-MQ resources that are enlisted in an XA transaction.
There is already a solution within A-MQ to migrate messages on scale down and this may be sufficient but it does need verifying and/or documenting.