Reflex 0.9: self.router.url.query not refreshed when URL is updated with window.history.replaceState

[TianyiLi10]

Describe the bug
I have a select filter that updates the URL query parameters whenever the selected filter value changes. The URL is updated using rx.call_script with window.history.replaceState.

Before upgrading to Reflex 0.9, updating the URL this way would also update the route query values accessible through self.router.url.query.

After upgrading to Reflex 0.9, the browser URL changes correctly, but self.router.url.query no longer reflects the updated query parameters.

I run with reflex run --env prod

To Reproduce
Steps to reproduce the behavior:

1. Create a select filter component, for instance with query parameter: name.
2. On filter change, update the URL query parameters using:
   rx.call_script(
   "window.history.replaceState({}, '', '?name=test')"
   )
3. Read the current query parameters from: self.router.url.query
4. Observe that after the URL is updated, self.router.url.query still contains the old values.

• Code/Link to Repo:

Expected behavior
When the URL query parameters are modified via window.history.replaceState, the router state should stay in sync and self.router.url.query should reflect the updated query parameters.

Screenshots
If applicable, add screenshots to help explain your problem.

Specifics (please complete the following information):

• Python Version: 3.11
• Reflex Version: 0.9.4
• OS: Linux
• Browser (Optional): Chrome

Additional context
Add any other context about the problem here.

[linear-code]

ENG-9685

[masenf]

@TianyiLi10 when you call the replaceState are you expecting the on_load events for the page to fire again with the new URL? Or just update the URL and get it on the backend the next time an event fires?

[masenf]

The PR #6625 has the following behavior:

• calling replaceState or pushState with an rx.call_script will not in itself trigger any reactive update or on_load to occur. This can be thought of as a "cosmetic" URL change. If you want this to reactively update vars that depend on the URL, but do NOT want on_load to trigger, then specify a no op callback event in your actual state for the rx.call_script, which will be sent with the new URL values.
• using rx.redirect(..., replace=True) will replace the URL (so no new history entry), but it does it via a navigate call, so it WILL fire on_load and the new URL will be reactively sent to the backend and process any computed var updates.

I think this is a reasonable compromise, exposes options for both types of behavior and keeps the overall code change minimal.