[ci] e2e_tests seems to be flaky

[ditman]

It seems that the e2e_tests.sh script is flaky, for example:

• https://github.com/google/A2UI/actions/runs/25881317555/job/76061759428#step:9:87
• https://github.com/google/A2UI/actions/runs/25834367642/attempts/1
• https://github.com/google/A2UI/actions/runs/25758114131/job/75651870151#step:9:32
• Maybe more in search (?)

Anyway, this is turning the tree red at seemingly random moments. This looks like an LLM capacity issue.

We need to fix/alleviate this in order to start enforcing "do not land on red" policies.

[polina-c]

Statuses for test run:

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/collaborating-on-repositories-with-code-quality-features/about-status-checks#check-statuses-and-conclusions

[ditman]

Statuses for test run:

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/collaborating-on-repositories-with-code-quality-features/about-status-checks#check-statuses-and-conclusions

This seems unusable directly from github actions: https://github.com/orgs/community/discussions/9875

[ditman]

Polina has a fix to make the affected test in an hourly schedule, rather than on post-submit:

• Update e2e_test.yaml to run on schedule, because it is flacky. #1441