PR created by the Librarian CLI to initialize a release. Merging this PR
will auto trigger a release.
Librarian Version: v0.7.0
Language Image:
us-central1-docker.pkg.dev/cloud-sdk-librarian-prod/images-prod/python-librarian-generator@sha256:b8058df4c45e9a6e07f6b4d65b458d0d059241dd34c814f151c8bf6b89211209
<details><summary>google-cloud-bigtable: 2.35.0</summary>
##
[2.35.0](v2.34.0...v2.35.0)
(2025-12-16)
### Features
* add basic interceptor to client (#1206)
([6561cfa](6561cfac))
* Add encodings for STRUCT and the Timestamp type
([72dfdc4](72dfdc44))
* add PeerInfo proto in Bigtable API
([72dfdc4](72dfdc44))
* Add Type API updates needed to support structured keys in materialized
views
([72dfdc4](72dfdc44))
* support mTLS certificates when available (#1249)
([ca20219](ca20219c))
### Bug Fixes
* re-export AddToCell for consistency (#1241)
([2a5baf1](2a5baf11))
* async client uses fixed grace period (#1236)
([544db1c](544db1cd))
* Deprecate credentials_file argument
([72dfdc4](72dfdc44))
* Add ReadRows/SampleRowKeys bindings for materialized views
([72dfdc4](72dfdc44))
* retry cancelled errors (#1235)
([e3fd5d8](e3fd5d86))
</details>
Previously, when a channel refresh occurs, the async client would use channel.close() with a grace parameter to allow previous channels to keep serving old requests for a time. We were seeing flakes in our tests, showing that channel.close() isn't reliable, and can sometimes cancel ongoing requests before the grace period ends
This PR fixes this by using a fixed sleep time before calling close in the async client, like the sync client already does. This should remove the potential for cancelled requests before the grace period ends, and improve test flakiness
I also updated the system test to fully capture this problematic state, instead of encountering it in a random race condition