Self-hosted Julia 1 CI jobs intermittently lose runner communication

Several Julia 1 CI jobs on the self-hosted 4 vCPU / 8 GB runner fail without Julia test logs. The check annotation is:\n\n> The self-hosted runner lost communication with the server. Verify the machine is running and has a healthy network connection. Anything in your workflow that terminates the runner process, starves it for CPU/Memory, or blocks its network access can cause this error.\n\nObserved jobs:\n\n- PR #307, commit 36fd53c6a407827efc7697493415382007ead52e: Core Julia 1 failed after 47m2s, job 86726935169. Datafit Julia 1 passed in the same run after reducing the Bayesian test workload.\n- Main after #306 merge, commit 6624ad45d2a9c44f9909ec981e4a1a7880be54a0: Core job 86714351300 and Datafit job 86714351296 both failed with the same runner-loss annotation.\n- Main after #305, commit b7e28a4766a61acd1b2c03841da4b6681ab0c3c8: Datafit job 86283697299 failed with the same runner-loss annotation while Core passed.\n- Main after #304, commit cea212159dfb540de6dff913ce3e94d99e95de19: Core job 86264617991 and Datafit job 86264618004 failed with the same runner-loss annotation.\n\nThe current PR #307 removes the known Datafit pressure point enough for Datafit to pass on CI, but Core still fails through the infrastructure/runner path rather than a Julia assertion. This likely needs runner/resource handling or CI partitioning rather than a package-code assertion fix.

No source file or test is named. Start by reviewing the GitHub Actions workflow and the self-hosted runner's resource and network logs, then compare the listed Core and Datafit job failures. Done means the affected Julia 1 jobs complete reliably on the runner and produce their expected test logs.