[Copilot]

This revision removes the previously introduced connectedCallback microtask batching and keeps only the high-value safety fix: a single element error should not abort a dateObserver tick for all observed elements. It also drops per-tick locale/time-zone/hour-cycle caching from this branch to avoid correctness risk and overlap with existing global Intl caching.

• Scope reset: restore main lifecycle semantics

  • Removed module-level pending render queue (pendingElements, pendingFlush, flushPending).
  • Restored synchronous connectedCallback() render (this.update()).
  • Restored disconnectedCallback() and attributeChangedCallback() behavior to match main (no pending-queue coupling).

• Observer robustness: per-element error isolation

  • Guarded each timeEl.update() inside dateObserver.update() so one throw does not stop remaining elements or prevent rescheduling.
  • Preserved global error visibility by rethrowing asynchronously.

• Tests updated to lock intended behavior

  • Removed the full connectedCallback microtask batching suite.
  • Added synchronous-render contract coverage (text/title present immediately after append).
  • Added error-isolation coverage demonstrating other elements continue updating and observer scheduling continues when one element throws.

for (const timeEl of this.elements) {
  nearestDistance = Math.min(nearestDistance, getUnitFactor(timeEl))
  try {
    timeEl.update()
  } catch (error) {
    setTimeout(() => {
      throw error
    })
  }
}

[mattcosta7]

Superseded by #399, which isolates errors per element during a shared-timer tick on top of #396. It also covers getUnitFactor (a throwing date getter, or an invalid time-zone on an element an hour or more old), which ran outside this PR's try. It checks that each error is reported exactly once, and it's based on current main instead of a branch 61 commits behind.