Environment:
- iOS App Version: 2026.2.1 (Build 2026.1766)
- Home Assistant Server Version: 2026.3.2
- Device: iPhone 14 (iPhone14,5)
- iOS Version: 26.3.1 (23D771330a)
Description:
When scanning an NFC tag that triggers a scan_tag webhook, the Home Assistant iOS app is killed by iOS due to exceeding the per-process memory limit. This happens reliably and reproducibly — even when the app is fully loaded in the foreground before scanning the tag.
The issue can also be reproduced by calling the NFC scan URL handler directly (without a physical NFC tag).
Steps to Reproduce:
- Open the Home Assistant app and wait until it is fully loaded
- Scan an NFC tag configured to trigger a
scan_tag webhook (or call the equivalent URL handler directly)
- The app is terminated by iOS immediately after the webhook is sent
Expected Behavior:
The app remains open after the NFC scan and the automation is triggered normally.
Actual Behavior:
The app is killed by iOS. The Jetsam log identifies the kill reason as per-process-limit, meaning the app exceeded its allocated memory budget.
Root Cause Analysis:
The Jetsam event log (bug_type: 298) clearly identifies the Home Assistant app as the largestProcess and shows it was killed due to per-process-limit:
"largestProcess" : "Home Assistant",
"name" : "Home Assistant",
"reason" : "per-process-limit",
"lifetimeMax" : 134272,
"pid" : 1008
With a page size of 16,384 bytes, lifetimeMax of 134,272 pages corresponds to approximately 2.05 GB of memory — the hard ceiling iOS enforces for this device class.
The app log shows that immediately after the NFC scan triggers the app, a full subscribe_entities request is initiated (WebSocket Request ID 13), loading the complete entity state for a large Home Assistant instance:
[HAAPI.swift:844] UpdateSensors > updating sensors [...]
[AppDatabaseUpdater.swift] Successfully saved 1625 entity registry entries
[AppDatabaseUpdater.swift] Successfully saved 197 device registry entries
[AppDatabaseUpdater.swift] ✅ Full update for server Home completed in 0.906s
This full entity reload — 1,625 entity registry entries and 197 device registry entries — appears to push memory consumption over the iOS limit, causing the process to be killed.
Additional context from app log:
The app log also shows a recurring error on every startup that may be related:
[AppDelegate.swift:312] checkForAlerts() > check error:
The first or last element was requested for an empty sequence
And a WebSocket subscription inconsistency:
[Environment.swift:76] WebSocket: unable to find request for identifier HARequestIdentifier(rawValue: 6)
[Environment.swift:76] WebSocket: unable to find subscription for identifier HARequestIdentifier(rawValue: 6)
Jetsam memory state at time of kill:
"memoryStatus": {
"compressorSize": 41704,
"compressions": 1997039,
"decompressions": 1258289,
"uncompressed": 211946,
"memoryPages": {
"active": 62280,
"inactive": 56799,
"fileBacked": 46250,
"wired": 56753,
"anonymous": 77872,
"free": 7043
}
}The system was already under memory pressure (only 7,043 free pages ≈ ~112 MB free) when the kill occurred.
Suspected fix area:
The full entity reload triggered by an NFC scan (or any URL-handler-based app activation) should either be deferred, streamed incrementally, or avoided entirely if the local database is already up to date. Triggering a complete registry refresh in response to a simple NFC scan seems disproportionate and is likely the proximate cause of the memory spike.
Environment:
Description:
When scanning an NFC tag that triggers a
scan_tagwebhook, the Home Assistant iOS app is killed by iOS due to exceeding the per-process memory limit. This happens reliably and reproducibly — even when the app is fully loaded in the foreground before scanning the tag.The issue can also be reproduced by calling the NFC scan URL handler directly (without a physical NFC tag).
Steps to Reproduce:
scan_tagwebhook (or call the equivalent URL handler directly)Expected Behavior:
The app remains open after the NFC scan and the automation is triggered normally.
Actual Behavior:
The app is killed by iOS. The Jetsam log identifies the kill reason as
per-process-limit, meaning the app exceeded its allocated memory budget.Root Cause Analysis:
The Jetsam event log (
bug_type: 298) clearly identifies the Home Assistant app as thelargestProcessand shows it was killed due toper-process-limit:With a page size of 16,384 bytes,
lifetimeMaxof 134,272 pages corresponds to approximately 2.05 GB of memory — the hard ceiling iOS enforces for this device class.The app log shows that immediately after the NFC scan triggers the app, a full
subscribe_entitiesrequest is initiated (WebSocket Request ID 13), loading the complete entity state for a large Home Assistant instance:This full entity reload — 1,625 entity registry entries and 197 device registry entries — appears to push memory consumption over the iOS limit, causing the process to be killed.
Additional context from app log:
The app log also shows a recurring error on every startup that may be related:
And a WebSocket subscription inconsistency:
Jetsam memory state at time of kill:
The system was already under memory pressure (only 7,043 free pages ≈ ~112 MB free) when the kill occurred.
Suspected fix area:
The full entity reload triggered by an NFC scan (or any URL-handler-based app activation) should either be deferred, streamed incrementally, or avoided entirely if the local database is already up to date. Triggering a complete registry refresh in response to a simple NFC scan seems disproportionate and is likely the proximate cause of the memory spike.