Redeploy removes prior asset bundles, breaking open sessions

[dcnsakthi]

Problem statement

Each deploy emits new content-hashed bundles and the previous ones stop being served. A browser tab that was open before the deploy will request a chunk that no longer exists and fail, with no built-in recovery.

Evidence. Cache headers are correct, so this is not a caching bug. See Checked and found working. The failure mode is asset lifetime, not cache policy. We did not manage to reproduce the blank page during this round of work, so this needs a clean repro before filing.

Impact. In an app deployed during working hours, users with an open tab hit a broken lazy load until they reload.

Proposed solution

Retain the previous deployment's assets for a grace period. Separately, document a client-side chunk-load-error handler that triggers a reload, since that pattern is needed regardless.

Alternatives considered

No response

Additional context

No response

[kdubau]

Thanks — good analysis, and agreed this is asset lifetime rather than caching. Preserving app availability across a redeploy (retaining prior bundles for a grace period) is tracked on our side. A client-side chunk-load-error handler that triggers a reload is a solid stopgap in the meantime. We will update here as this progresses.