FunctionsNetHost exits with code 139 (SIGSEGV) on .NET 10 isolated worker at normal operating load on Flex Consumption (Linux)

### Description

On **.NET 10 isolated worker** running on **Flex Consumption (Linux)**, `FunctionsNetHost` crashes with exit code **139 (0x8B)** — a SIGSEGV (segmentation fault). The crashes occur at **normal, routine production load** — not under stress conditions — and repeat in rapid succession, destabilizing the function app for 6–10 minute windows.

The configured concurrency at time of crash:
- **`maxConcurrentSessions: 3`** — the Service Bus session processor ceiling, not a peak
- **`maxConcurrentCalls: 16`**
- **`maxMessageBatchSize: 1`**

No application code exception precedes the crash. The worker process simply terminates with SIGSEGV. The Azure Functions HOST process (`HostInstanceId` is stable throughout) detects the crash and restarts the worker, but under continued normal load the worker crashes again within 1–2 minutes.

The crashes affect **two separate function apps** in the same environment on the same day.

---

## Exception / Error Log

### GroupUpdaterFunction (GraphUpdater function app)

```
Exception while executing function: Functions.GroupUpdaterFunction
Microsoft.Azure.WebJobs.Host.FunctionInvocationException
---> Microsoft.Azure.WebJobs.Script.Workers.WorkerProcessExitException:
/azure-functions-host/workers/dotnet-isolated/bin/FunctionsNetHost exited with code 139 (0x8B)
---> System.Exception: n/a

at Microsoft.Azure.WebJobs.Script.Workers.WorkerProcess.ThrowIfExitError
(WorkerProcess.cs:233) [Microsoft.Azure.WebJobs.Script.Grpc, Version=4.1047.0.0]
at Microsoft.Azure.WebJobs.Script.Workers.WorkerProcess.OnProcessExited
(WorkerProcess.cs:185) [Microsoft.Azure.WebJobs.Script.Grpc, Version=4.1047.0.0]
at Microsoft.Azure.WebJobs.Script.Description.WorkerFunctionInvoker+d__9.MoveNext
(WorkerFunctionInvoker.cs:103) [Microsoft.Azure.WebJobs.Script.Grpc, Version=4.1047.0.0]
```

### DeltaLinkUploaderFunction → SubOrchestratorFunction → OrchestratorFunction (GroupMembershipObtainer function app)

```
Microsoft.Azure.WebJobs.Host.FunctionInvocationException:
Exception while executing function: Functions.SubOrchestratorFunction
---> DurableTask.Core.Exceptions.OrchestrationFailureException:
Task 'DeltaLinkUploaderFunction' (#8) failed with an unhandled exception:
Exception while executing function: Functions.DeltaLinkUploaderFunction
[Microsoft.Azure.WebJobs.Extensions.DurableTask, Version=3.0.0.0]
```

---

## Crash Events

### Event 1 — GraphUpdater function app

**Date/time window:** 2026-03-05, 16:53 – 17:02 UTC
**Worker restarts:** 16:53:01 · 16:55:41 · 16:57:01 · 16:59:51 · 17:00:41 · 17:01:41 UTC
**Duration of instability:** ~9 minutes, 6 worker crashes

| Identifier | Value |
|---|---|
| `HostInstanceId` | `b83e8852-05dd-4c8b-9631-4b23867e7342` (stable throughout — HOST did not crash) |
| DTF Orchestration RunId | `bfac744d-0c12-5729-8d64-7f6be374114b` |

### Event 2 — GroupMembershipObtainer function app

**Date/time:** 2026-03-05, 23:20 – 23:21 UTC
Two concurrent compute instances affected simultaneously.

**Instance A** — DeltaLinkUploaderFunction → SubOrchestratorFunction crashed

| Identifier | Value |
|---|---|
| Timestamp | `2026-03-05T23:20:54.4169845Z` |
| `HostInstanceId` | `2bdf0b18-35f9-4e87-a511-bbeeb759df7a` |
| `cloud_RoleInstance` | `0--de8f0470-1b46-41c7-89de-3c813d91e906` |
| `ProcessId` | `34` |
| `InvocationId` (SubOrchestratorFunction) | `4d3ae322-a843-4003-9c16-41e5590c1ead` |
| `operation_Id` | `7ce913159e53623eb759b626fd6314f6` |

**Instance B** — OrchestratorFunction failure propagated from same root crash

| Identifier | Value |
|---|---|
| Timestamp | `2026-03-05T23:20:59.7993100Z` |
| `HostInstanceId` | `efe7cf84-3f6f-4418-acc2-07d1cb9096c0` |
| `cloud_RoleInstance` | `0--c133ae49-0590-43a4-ac05-a30950dd3ce4` |
| `ProcessId` | `34` |
| `InvocationId` (OrchestratorFunction) | `1c1ae40f-af9a-499d-adc3-0853781db8e1` |
| `operation_Id` | `2afc24c3598368fc87747be9de4b323f` |

### Event 3 — GraphUpdater function app (same day, later)

**Date/time:** 2026-03-05T23:26:49.9042201Z

| Identifier | Value |
|---|---|
| Timestamp | `2026-03-05T23:26:49.9042201Z` |
| `HostInstanceId` | `b99961179-d28e-48b3-aabf-65ac5d9e1f54` |
| `ProcessId` | `35` |
| `InvocationId` (GroupUpdaterFunction) | `0520f145-bc19-4004-bc8d-a0fa8611b768` |
| Invocation duration at time of crash | `46,155 ms` |

---

## Steps to Reproduce

Not deterministically reproducible on demand, but the confirmed runtime conditions at time of all crashes:

1. Flex Consumption plan, Linux, .NET 10 isolated worker
2. Durable Functions active (`extendedSessionsEnabled: true`, 30s idle timeout, `maxConcurrentSessions: 3`)
3. Service Bus session-based trigger processing at its configured ceiling (`maxConcurrentSessions: 3`)
4. Under these routine operating conditions, `FunctionsNetHost` crashes with SIGSEGV — no exception in application code precedes it

---

## Expected Behavior

The worker process should not crash with a segmentation fault under routine production load. If a crash does occur, it should not recur repeatedly within the same operating window.

---

## Actual Behavior

`FunctionsNetHost` crashes with exit code 139 (0x8B / SIGSEGV) repeatedly. The Azure Functions host detects the crash, restarts the worker, but under continued normal load the worker crashes again within 1–2 minutes. The cycle repeats 6 times over ~9 minutes before stabilizing.

---

## Downstream Impact: Durable Entity State Loss

Because `FunctionsNetHost` terminates mid-execution, in-flight Durable Entity signals are orphaned — the entity dispatch is sent to the worker but the worker crashes before completing, leaving no `ContinuedAsNew` response written to Table Storage.

Observed sequence (Event 1, all times UTC):

| Time | Sequence | Event |
|---|---|---|
| `2026-03-05T16:56:02Z` | seq 1726 | Entity signal dispatched → **no completion log** (orphaned dispatch) |
| `2026-03-05T16:56:05Z` | seq 1731 | Entity state committed via `ContinuedAsNew:True` with `{MessagesProcessed:3, IsValidGroup:true}` |
| `2026-03-05T16:57:06Z` | seq 1851 | Entity re-executed in **~1ms** — `InitializeState()` called (fresh/null state), not loaded from Table Storage — 61 seconds after confirmed write |

The state loss cascades: only 2 of 5 expected `MessagesProcessed` increments are counted → business-level sync job failure ("Not all messages were processed, only 2 out of 5").

---

## Environment

| Property | Value |
|---|---|
| OS | Linux |
| Azure Region | `westus2` |
| Hosting plan | Flex Consumption |
| Worker model | Isolated (out-of-process) |
| Worker target framework | `net10.0` |
| .NET SDK (global.json) | `10.0.103` (`rollForward: latestFeature`) |
| .NET Runtime | `10.0.3` |
| Azure Functions host SDK | `azurefunctions: 4.1047.100.26071` |
| `Microsoft.Azure.WebJobs.Script.Grpc` | `4.1047.0.0` |
| `Microsoft.Azure.WebJobs.Script` | `4.1047.0.0` |
| `Microsoft.Azure.WebJobs.Host` | `3.0.44.0` |
| `Microsoft.Azure.WebJobs.Extensions.DurableTask` (host-side) | `3.0.0.0` |
| `Microsoft.Azure.Functions.Worker` (NuGet) | `2.51.0` |
| `Microsoft.Azure.Functions.Worker.Sdk` (NuGet) | `2.0.7` |
| `Microsoft.Azure.Functions.Worker.Extensions.DurableTask` (NuGet) | `1.16.0` |
| `Microsoft.DurableTask.Abstractions` (NuGet) | `1.22.0` |
| `maxConcurrentSessions` | `3` (configured ceiling, not a stress peak) |
| `maxConcurrentCalls` | `16` |
| `extendedSessionsEnabled` | `true` (30s idle timeout) |
| `functionTimeout` | `00:20:00` |

---

## Additional Context

[host.json](https://github.com/user-attachments/files/25805576/host.json)

[crash-instance-1-20260305-232054-unknown-op-sanitized.json](https://github.com/user-attachments/files/25805614/crash-instance-1-20260305-232054-unknown-op-sanitized.json)
[crash-instance-2-20260305-232054-unknown-op-sanitized.json](https://github.com/user-attachments/files/25805613/crash-instance-2-20260305-232054-unknown-op-sanitized.json)
[crash-instance-2-20260305-232059-unknown-op-sanitized.json](https://github.com/user-attachments/files/25805616/crash-instance-2-20260305-232059-unknown-op-sanitized.json)
[crash-instance-3-20260305-232059-unknown-op-sanitized.json](https://github.com/user-attachments/files/25805615/crash-instance-3-20260305-232059-unknown-op-sanitized.json)

- The HOST process remains stable throughout all crashes (constant `HostInstanceId` per event) — only `FunctionsNetHost` (the dotnet-isolated worker) crashes
- The crash is a raw SIGSEGV with no managed exception preceding it — no application-level exception handler fires before the process terminates
- The crashes affect **two independent function apps** in the same environment
- **Three separate crash events** occurred on 2026-03-05, spanning two function apps across two time windows (16:53–17:02 UTC and 23:20–23:27 UTC)
- Also observed on 2026-03-06
- Subscription ID and function app resource ID can be provided privately on request — not included here to avoid exposing sensitive identifiers publicly