Selenium Grid 4.34 — Parallel Session Isolation Issue After Upgrade

[Stan]

Hello everyone,
We recently upgraded our Selenium Grid deployment (running in Docker) from 4.32 to 4.34 and encountered a major regression in parallel execution.

Current setup:
12 Grid nodes
maxSessions=10 per node
SE_NODE_OVERRIDE_MAX_SESSIONS=true (enabled)
Each node provisioned with 6 CPU and 6 GB RAM
Using Docker images from SeleniumHQ
Observed behavior after 4.34 upgrade:
High rate of TimeoutException and ElementClickInterceptedException errors
Symptoms suggest parallel sessions are not fully isolated — locators or click actions appear to be “bleeding” between tests, as if multiple sessions share the same browser instance/thread state

Behavior is reproducible and consistent even if I set 3 replicas with 3 max sessions 3cpu and 3gb each and run 6 tests in parallel.

Workaround:
Switching to 15 replicas with:
maxSessions=1
1 CPU and 1 GB per node
No override of maxSessions
Restores stability — tests pass consistently.

Question:
I’ve seen in the Selenium docs that 1 CPU / 1 GB / 1 session per node is the recommended setup, but was this recommendation enforced in 4.34?
It seems that in previous versions, maxSessions>1 worked without isolation issues, but now session concurrency per node may no longer be safe or officially supported.

Has anyone else experienced the same isolation/thread-safety problems with multiple sessions per node after upgrading to 4.34? Was there a breaking change related to session management or ChromeDriver handling in this release?