On-demand is a stored file the player can fetch ahead of playback, so a slow moment gets absorbed and you never see it. Live has no such freedom: segments of roughly six seconds are produced as the event happens, and the player can only hold a short cushion. Any interruption longer than that cushion becomes a visible freeze. The two also frequently travel different routes, so a path that is congested, filtered or blocked for live can be perfectly fine for on-demand. That is why on-demand playing normally is genuine evidence rather than a coincidence: it clears your credentials, your app install and your line in a single test.

1. 1On-demand can buffer ahead, live cannot
2. 2The two often use different delivery paths
3. 3Working on-demand clears the account, the app and the line

Open three live channels from three different groups before you decide anything. If every live channel fails, look at the session and the network path: a connection slot in use elsewhere, a restrictive network filtering the live path, or a stale session after standby. If one whole group is dark while other groups play, that is server-side and belongs in a support ticket with the group name and the time. If a single channel is black, that source has dropped at origin. Three symptoms, three owners, and only the first of them is even partly yours to fix from the sofa.

1. 1All live dead: session, slot or network path
2. 2One group dark: server-side, report it with the time
3. 3One channel black: source down at origin

It looks exactly like a slow connection, which is why so many people replace routers that were never at fault. Run the test that separates them: on a frozen live channel, switch away, wait five seconds, and switch back. If the picture returns immediately and holds, the session stalled and a new one repaired it, so bandwidth was never the constraint. Stalls are common after a device wakes from standby, because the app resumes with a streaming handle the server already dropped while the interface still looks normal from cached data. If switch-back keeps fixing it, force-close the app before the device sleeps.

1. 1Instant recovery on switch-back means a stalled session
2. 2Standby and wake is the usual trigger
3. 3Force-close before sleep if the pattern repeats

Because everyone tunes in at once. Live sport is the point at which viewers judge a service and leave it, and the failure they describe is not gradual decline but a freeze during something they cared about. Mechanically it is concurrency: many subscribers on the same path at the same minute, arriving faster than the capacity provisioned for them. That is not a device problem and no setting reaches it, which is worth saying rather than blaming a router. It is the reason published uptime deserves weight in a purchase decision. IP4KTV publishes 99.99%, about 53 minutes of downtime across a year, and a $5 24-hour trial lets you test a busy evening.

1. 1Peak concurrency is the mechanism, not your hardware
2. 2The moment people judge a service is a live event
3. 3Test a busy evening before paying for months