You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
🗣️ Enable betaEnableViewUpdates by default — opt-in period #4974
We are enabling partial view updates by default in an upcoming minor release. Until then, the feature is available behind the betaEnableViewUpdates option for early adopters to try and report issues.
What it does
When enabled, the RUM SDK sends only the fields that changed on intermediate view updates instead of re-sending the full view event every time. This reduces bandwidth and intake volume with no loss of information.
How to try it
Set betaEnableViewUpdates: true in your datadogRum.init() configuration:
Now: opt-in period open. Please try it on staging/production traffic and report any anomaly (missing fields, broken views in the UI, unexpected errors).
~2 weeks: we flip the default to true in a minor release. The option will remain to disable the feature if needed.
What to look for
Views rendering correctly in the RUM explorer (durations, error/long-task/resource metrics).
Any unexpected errors or anomalies in the console.
Bandwidth/intake volume reduction (expected, but worth confirming).
Feedback
Reply to this issue with:
✅ Works fine for us (version, app type, traffic volume)
changed the title [-]Enable `betaEnableViewUpdates` by default — opt-in period[/-][+]🗣️ Enable `betaEnableViewUpdates` by default — opt-in period[/+]on Aug 24, 2026
❌ After upgrading @datadog/browser-rum from 7.14.0 to 7.15.0 on October 7, 2026, we’re seeing many more views without page-load timings. We use npm with site: 'ddog-gov.com', no proxy, and no explicit betaEnableViewUpdates setting.
Observed impact
We compared initial-load views from October 1–12 in the same application and DEMO environment:
Chrome traffic
Initial loads
Missing LCP
Versions before the upgrade
7,119
10.2%
Versions after the upgrade
3,627
82.4%
We counted views with and without the LCP field. HeadlessChrome also had 81.0% missing LCP after the upgrade. The increase within ordinary Chrome means a larger headless traffic share alone doesn’t explain it, although Chrome can also be automated.
We also found views missing FCP, loading time, first byte and DOM/load-event timings. We haven’t measured their missing percentages separately.
Real affected event
This redacted excerpt comes from RUM Explorer’s JSON. Durations are in nanoseconds:
The full view object has none of the page-load timing fields listed above. It is marked ended and records foreground activity throughout its 5.456-second duration.
Partial-update investigation
In a separate, local reproduction, we intercepted outgoing requests. With 7.15.0 defaults, later timings appeared in view_update events. With betaEnableViewUpdates: false, they appeared in full view events again.
That confirms the payload change. We haven’t yet verified that disabling the option restores timings in our live org.
The timing of the increase and the reproduction suggest a connection to partial view updates, but we don’t know where the failure occurs. Could you help trace an affected view through intake, merging and Explorer retrieval?
Hey @seansusmilch, thank you for the report. I'm going to investigate right now, in the meantime, can I suggest you to set betaEnableViewUpdates: false as you've done before? That should bring back those metrics.
Activity
[-]Enable `betaEnableViewUpdates` by default — opt-in period[/-][+]🗣️ Enable `betaEnableViewUpdates` by default — opt-in period[/+]seansusmilch commentedon Oct 13, 2026
❌ After upgrading
@datadog/browser-rumfrom 7.14.0 to 7.15.0 on October 7, 2026, we’re seeing many more views without page-load timings. We use npm withsite: 'ddog-gov.com', no proxy, and no explicitbetaEnableViewUpdatessetting.Observed impact
We compared initial-load views from October 1–12 in the same application and DEMO environment:
We counted views with and without the LCP field. HeadlessChrome also had 81.0% missing LCP after the upgrade. The increase within ordinary Chrome means a larger headless traffic share alone doesn’t explain it, although Chrome can also be automated.
We also found views missing FCP, loading time, first byte and DOM/load-event timings. We haven’t measured their missing percentages separately.
Real affected event
This redacted excerpt comes from RUM Explorer’s JSON. Durations are in nanoseconds:
{ "type": "view", "view": { "id": "REDACTED", "loading_type": "initial_load", "is_active": false, "time_spent": 5456000000, "in_foreground_periods": [ { "start": 0, "duration": 5456000000 } ], "cumulative_layout_shift": 0 } }The full
viewobject has none of the page-load timing fields listed above. It is marked ended and records foreground activity throughout its 5.456-second duration.Partial-update investigation
In a separate, local reproduction, we intercepted outgoing requests. With 7.15.0 defaults, later timings appeared in
view_updateevents. WithbetaEnableViewUpdates: false, they appeared in fullviewevents again.That confirms the payload change. We haven’t yet verified that disabling the option restores timings in our live org.
The timing of the increase and the reproduction suggest a connection to partial view updates, but we don’t know where the failure occurs. Could you help trace an affected view through intake, merging and Explorer retrieval?
mormubis commentedon Oct 13, 2026
Hey @seansusmilch, thank you for the report. I'm going to investigate right now, in the meantime, can I suggest you to set
betaEnableViewUpdates: falseas you've done before? That should bring back those metrics.Thanks!
seansusmilch commentedon Oct 15, 2026
Hey @mormubis, thanks for taking a look at this! Do you have any insight as to why these metrics go missing with the partial view updates?
We ended up rolling back the package yesterday and that seemed to bring back coverage to near what we had before.
Next we will upgrade to 7.15.0 again along with disabling the
betaEnableViewUpdatesflag