INP metric: Too many spans with <unknown> target due to disappearing DOM elements
Problem Statement
We are seeing a large number of INP-related spans where the clicked element is reported as <unknown>.
This happens when the user clicks UI elements that immediately disappear from the DOM right after the interaction — most commonly:
- closing buttons (e.g., closing a modal),
- clearing input with an “X” button,
- removing items from lists, cards, etc.
Because these elements are removed, by the time Sentry processes the INP event, e.target is already null, which results in spans showing <unknown> instead of the actual clicked element.
This significantly reduces the usefulness and traceability of INP measurements.
Solution Brainstorm
A potential fix is to capture and cache metadata about the clicked element at the moment of the click event — before the element disappears.
Example approach:
- On click, immediately compute and store the result of
htmlTreeAsString(event.target)(or an equivalent helper used for INP). - When constructing INP spans, use the cached
htmlTreeAsStringvalue instead of trying to readevent.targetagain (which may already be null after DOM removal).
Product Area
Unknown
Assigning to @getsentry/support for routing ⏲️
Routing to @getsentry/product-owners-insights for triage ⏲️
What is the SDK version used here? We shipped a fix in 10.23.0 that should reduce the number of these elements considerably as we observed it on our own end.
What is the SDK version used here? We shipped a fix in 10.23.0 that should reduce the number of these elements considerably as we observed it on our own end.
- @sentry/react-native: 7.7.0
- @sentry/core: 10.26.0
- @sentry/browser: 10.26.0
- @sentry/react: 10.26.0
- @sentry/webpack-plugin: 4.6.0
In our project: over the last 14 days, approximately 50% of the traces named ui.interaction.click have the span.description value set to <unknown>.
Perhaps the timing window we limit for the elements cache is a bit aggressive, I will see about increasing it and observing the results.