some time today, the server failed again, and when I tried to view a TV channel live, I got the usual scrambled mess. The PC was booted up about 1100BST, so it failed about 8 hours later.
https://i.imgur.com/TLjTjYm.png
I looked in /var/log/syslog and found this:
Sep 2 19:12:29 mcintyre kernel: [39238.047519] irq 18: nobody cared (try booting with the "irqpoll" option)
Sep 2 19:12:29 mcintyre kernel: [39238.047549] CPU: 0 PID: 0 Comm: swapper/0 Tainted: G OE 4.4.0-116-lowlatency #140-Ubuntu
Sep 2 19:12:29 mcintyre kernel: [39238.047551] Hardware name: /DH61CR, BIOS BEH6110H.86A.0028.2011.0715.1815 07/15/2011
Sep 2 19:12:29 mcintyre kernel: [39238.047552] 0000000000000086 731cdad3b862bbf2 ffff88011fa03e60 ffffffff81409811
Sep 2 19:12:29 mcintyre kernel: [39238.047555] ffff8800d6469600 ffff8800d64696d4 ffff88011fa03e88 ffffffff810e33c3
Sep 2 19:12:29 mcintyre kernel: [39238.047557] ffff8800d6469600 0000000000000000 0000000000000012 ffff88011fa03ec0
Sep 2 19:12:29 mcintyre kernel: [39238.047559] Call Trace:
Sep 2 19:12:29 mcintyre kernel: [39238.047561] <IRQ> [<ffffffff81409811>] dump_stack+0x63/0x82
Sep 2 19:12:29 mcintyre kernel: [39238.047568] [<ffffffff810e33c3>] __report_bad_irq+0x33/0xc0
Sep 2 19:12:29 mcintyre kernel: [39238.047570] [<ffffffff810e3757>] note_interrupt+0x247/0x290
Sep 2 19:12:29 mcintyre kernel: [39238.047572] [<ffffffff810e0772>] handle_irq_event_percpu+0xc2/0x210
Sep 2 19:12:29 mcintyre kernel: [39238.047574] [<ffffffff810e08fc>] handle_irq_event+0x3c/0x60
Sep 2 19:12:29 mcintyre kernel: [39238.047577] [<ffffffff810e3dbc>] handle_fasteoi_irq+0x9c/0x190
Sep 2 19:12:29 mcintyre kernel: [39238.047580] [<ffffffff81032353>] handle_irq+0x23/0x30
Sep 2 19:12:29 mcintyre kernel: [39238.047583] [<ffffffff8185fd2b>] do_IRQ+0x4b/0xe0
Sep 2 19:12:29 mcintyre kernel: [39238.047585] [<ffffffff8185d3bf>] common_interrupt+0xbf/0xbf
Sep 2 19:12:29 mcintyre kernel: [39238.047586] <EOI> [<ffffffff816eae93>] ? cpuidle_enter_state+0x113/0x2f0
Sep 2 19:12:29 mcintyre kernel: [39238.047591] [<ffffffff816eb0a7>] cpuidle_enter+0x17/0x20
Sep 2 19:12:29 mcintyre kernel: [39238.047594] [<ffffffff810c9632>] call_cpuidle+0x32/0x60
Sep 2 19:12:29 mcintyre kernel: [39238.047596] [<ffffffff816eb089>] ? cpuidle_select+0x19/0x20
Sep 2 19:12:29 mcintyre kernel: [39238.047598] [<ffffffff810c991e>] cpu_startup_entry+0x2be/0x390
Sep 2 19:12:29 mcintyre kernel: [39238.047601] [<ffffffff8184fee9>] rest_init+0x89/0x90
Sep 2 19:12:29 mcintyre kernel: [39238.047605] [<ffffffff81f6402f>] start_kernel+0x496/0x4b7
Sep 2 19:12:29 mcintyre kernel: [39238.047607] [<ffffffff81f63120>] ? early_idt_handler_array+0x120/0x120
Sep 2 19:12:29 mcintyre kernel: [39238.047609] [<ffffffff81f63339>] x86_64_start_reservations+0x2a/0x2c
Sep 2 19:12:29 mcintyre kernel: [39238.047611] [<ffffffff81f63485>] x86_64_start_kernel+0x14a/0x16d
Sep 2 19:12:29 mcintyre kernel: [39238.047613] handlers:
Sep 2 19:12:29 mcintyre kernel: [39238.047623] [<ffffffff810e0980>] irq_default_primary_handler threaded [<ffffffffc0828fa0>] saa716x_budget_pci_irq [saa716x_tbs_dvb]
Sep 2 19:12:29 mcintyre kernel: [39238.047669] Disabling IRQ #18
Sep 2 19:12:59 mcintyre kernel: [39267.747308] demux_worker: called but nothing to do
Sep 2 19:12:59 mcintyre kernel: [39267.847322] demux_worker: called but nothing to do
Sep 2 19:13:19 mcintyre kernel: [39287.747212] demux_worker: called but nothing to do
Sep 2 19:13:19 mcintyre kernel: [39287.847226] demux_worker: called but nothing to do
Sep 2 19:14:09 mcintyre kernel: [39337.846980] demux_worker: called but nothing to do
and many more demux workers log entries.
tried restarting TVH, still didn't work. had to reboot. immediately everything working again just fine.
I'll haul a monitor over and check that all unnecessary hardware is disabled, but it doesn't look as if the card is sharing its interrupt with anything.