ObjectDisposedException in ManualResetEventSlim during runtime queries (latency-dependent race condition)

Environment

• Npgsql Version: 10.0.1
• .NET Version: 10.0
• PostgreSQL: Supabase (via pooler.supabase.com - PgBouncer Transaction Mode)
• OS: Windows 11

Description

We are experiencing ObjectDisposedException on ManualResetEventSlim.Reset() in NpgsqlConnector.ResetCancellation() during runtime database queries.

Critical finding: The error is latency-dependent and 100% reproducible:

• Low latency (5-20ms): Never fails (Azure EU → Supabase EU)
• High latency (100-300ms): Always fails (Turkey → Supabase EU)

This suggests a race condition where the timing window is only hit under high-latency conditions.

Stack Trace

System.ObjectDisposedException: Cannot access a disposed object.
Object name: 'System.Threading.ManualResetEventSlim'.
   at System.Threading.ManualResetEventSlim.Reset()
   at Npgsql.Internal.NpgsqlConnector.ResetCancellation()
   at Npgsql.NpgsqlCommand.ExecuteReader(Boolean async, CommandBehavior behavior, CancellationToken cancellationToken)
   at Npgsql.NpgsqlCommand.ExecuteDbDataReaderAsync(CommandBehavior behavior, CancellationToken cancellationToken)

Minimal Reproduction (without EF Core)

using Npgsql;

// Connection to Supabase via PgBouncer Transaction Mode pooler
var connectionString = "Host=aws-0-eu-central-1.pooler.supabase.com;Database=postgres;Username=postgres.xxx;Password=xxx;Port=5432;SSL Mode=Require;Multiplexing=false";

// Simulate high-latency environment or run from location with >100ms latency to database
await using var conn = new NpgsqlConnection(connectionString);
await conn.OpenAsync();

// Run multiple parallel queries to increase race condition probability
var tasks = Enumerable.Range(0, 10).Select(async _ =>
{
    await using var cmd = new NpgsqlCommand("SELECT 1", conn);
    await cmd.ExecuteScalarAsync();
});

await Task.WhenAll(tasks); // ObjectDisposedException occurs here under high latency

Frequency

• Every query under high-latency conditions (100-300ms)
• Never under low-latency conditions (5-20ms)
• Deterministic, not random

Connection Details

• Supabase Pooler: PgBouncer in Transaction Mode (pooler.supabase.com)
• Multiplexing: Disabled (Multiplexing=false)
• SSL: Required

Attempted Workarounds (None Worked)

Analysis

The error occurs in ResetCancellation() which calls ManualResetEventSlim.Reset(). Under high latency:

1. Query execution takes longer due to network delay
2. This appears to trigger Npgsql's internal cancellation/timeout mechanism
3. A race condition occurs: one thread disposes ManualResetEventSlim while another calls Reset()

The latency dependency explains why:

• Production (low latency) never hits the timing window
• Remote development (high latency) consistently hits it

Related Issues

• ObjectDisposedException in ManualResetEventSlim when running dotnet ef database update with .NET 10 and Npgsql 10.0.0 efcore.pg#3699 - Same stack trace but only during EF Core migrations (design-time)
• Npgsql 7.0.2 blocks threads NpgsqlConnector.PerformUserCancellation causing connection pool exhaustion #5032 - Previous ManualResetEventSlim threading issues in 7.0.2

Additional Context

This issue affects all runtime queries, not just EF Core migrations. The EF Core issue #3699 specifically states "non-design time activities function" - our finding expands the scope significantly.