I've tested on a PG16, the issue is indeed triggered with the replication blocked while the conflicting query is stuck in ClientWrite.
I've cleaned up the tests: I've created a dedicated PgProto (definitely open to suggestions for a better name...) module containing all the helpers to send and receive messages on a raw socket in 0001. 0002 contains the change with the test using the new PgProto module. The failure on windows I had was due to using the wrong username. I was using $USER variable env and I'm now getting the user from an existing psql session instead. Also, I think the DoingBlockingWrite variable was unnecessary? Calling ProcessRecoveryConflictInterrupts should be enough as I don't think QueryCancelHoldoffCount can be >0 when writing on a socket it should always be able to cancel the statement.
v03-0001-Add-PgProto-test-module-to-send-message-on-a-raw.patch
Description: Binary data
v03-0002-Accept-recovery-conflict-interrupt-on-blocked-wr.patch
Description: Binary data
