Matching service GetTaskQueue latency metrics is very large

After updating temporal from v1.13.0 to v1.21.2, this metrics increased dramatically. Other metrics don’t look suspicious.
At the same time, all our workflows/activity work without problems, there are no big delays. Has anyone experienced this?

histogram_quantile(0.95, sum(rate(client_latency_bucket{operation="MatchingClientGetTaskQueueUserData"}[5m])) by (operation, le))

histogram_quantile(0.95, sum(rate(service_latency_bucket{operation="GetTaskQueueUserData"}[5m])) by (operation, le))

have observed something similar at my end,
It might be some default duration for this task similar to PollActivityTaskQueue or PollWorkflowTaskQueue which is 1 minute.