fix: the behavior of subscription in pubsub when nodes are unstable (… #1008

Job	Run time
Lint	10s
Benchmark	2m 46s
Massive Cluster	5m 30s
NAT-ted Environments	1m 5s
Main (7.2, 3.3)	1m 14s
Profiling (single)	45s
Main (7.2, 3.3, compose.ssl.yaml)	1m 26s
IPS	3m 6s
Profiling (excessive_pipelining)	46s
Main (7.2, 3.3, hiredis)	1m 0s
Profiling (pipelining_in_moderation)	50s
Main (7.2, 3.3, hiredis, compose.ssl.yaml)	1m 34s
Profiling (original_mget)	47s
Main (7.2, 3.3, compose.replica.yaml, 2)	1m 7s
Profiling (emulated_mget)	51s
Main (test_cluster_down)	4m 20s
Main (test_cluster_broken, no, 6)	2m 5s
Main (8, 3.3, compose.valkey.yaml, 2)	1m 8s
Main (7.2, 3.2, compose.auth.yaml)	1m 5s
Main (7.0, 3.1)	1m 4s
Main (6.2, 3.0)	1m 7s
Main (5.0, 2.7)	1m 15s
Main (test_cluster_state, PrimaryOnly, compose.valkey.yaml, 8, 2, 9)	2m 7s
Main (test_cluster_state, Pooled, compose.valkey.yaml, 8, 2, 9)	2m 1s
Main (test_cluster_state, ScaleReadRandom, compose.valkey.yaml, 8, 2, 9)	2m 4s
Main (test_cluster_state, ScaleReadRandomWithPrimary, compose.valkey.yaml, 8, 2, 9)	2m 4s
Main (test_cluster_state, ScaleReadLatency, compose.valkey.yaml, 8, 2, 9)	2m 8s
Main (jruby)	1m 39s
Main (truffleruby)	1m 20s
Main (test_cluster_scale, Single, compose.scale.yaml, 8)	1m 14s
Main (test_cluster_scale, Pipeline, compose.scale.yaml, 8)	1m 15s
Main (test_cluster_scale, Transaction, compose.scale.yaml, 8)	1m 16s
Main (test_cluster_scale, PubSub, compose.scale.yaml, 8)	1m 42s
	53m 51s

Provide feedback