vBoxx
    Back to help center
    Known issues

    macOS Tahoe TCP networking bug after 49 days of uptime

    Active Apple bug that silently breaks new TCP connections on long-running Tahoe hosts.

    5 min read

    What goes wrong

    After 49 days 17 hours of uptime, a kernel timer in macOS Tahoe freezes. Closed TCP connections stop being released and slowly fill the outbound port pool. Once the pool is full the Mac can no longer open new TCP connections. Ping still works, so the host looks alive while being effectively offline for SSH, VNC and most application traffic.

    Affected systems

    • macOS Tahoe 26.0 through at least 26.4.1.
    • Earlier macOS versions (Sequoia 15.x and below) are not affected — the bug was introduced in 26.0.

    Symptoms

    If your host has crossed the 49-day threshold and the bug has triggered:

    • Inbound SSH and VNC connections stop working.
    • New outbound TCP connections fail or stay in SYN_SENT.
    • ICMP (ping) keeps working.
    • TIME_WAIT count climbs and never drops: 
      netstat -an | grep -c TIME_WAIT

    Why most users never hit it

    Anyone applying updates regularly rarely accumulates 49 days of uptime. On hosted Macs running as servers, however, that uptime is normal.

    The failure is also a two-stage event. The kernel timer freezes at day 49. Visible symptoms only appear once enough outbound connections have accumulated to exhaust the ~16 000-port pool. A Mac that mainly accepts inbound traffic (web server, light SSH use) burns ports slowly and may stay reachable for days or weeks after the freeze. It will fail eventually.

    If a Tahoe host has uptime over 49 days, reboot it preventatively — even if it currently looks fine.

    Detection

    Quickest checks:

    uptime                                   # days since last boot
netstat -an | grep -c TIME_WAIT          # high and climbing = suspect
sysctl net.inet.tcp.tcps_time_wait       # implementation specific, monitor trend

    Workaround

    Reboot every 30 days. The cleanest mitigation. Schedule a monthly reboot from the portal or from inside macOS:

    # Schedule a reboot every 1st of the month at 04:00
sudo pmset repeat shutdown M 04:00:00

    (Or restart with sudo shutdown -r now from a cron / launchd job.)

    If you can't tolerate the downtime, you can run with FileVault disabled and use the Customer API to trigger a hard reboot from CI on a schedule.

    Status

    We have reported this to Apple. Until a fix ships in a Tahoe point release, monthly reboots remain the official recommendation on vBoxx Mac hosts running Tahoe.