Three things at once. It ingests the source feed continuously. It encodes that feed into one or more bitrate ladders, since a 4K version at 15-25 Mbps and a 1080p version at 5-8 Mbps are separate encodes with separate costs. And it serves segments over HTTP to every viewer currently on that channel, updating a manifest file each time a new segment is ready. The outbound cost scales linearly with concurrent viewers: a thousand people watching one 4K channel is 15-25 Gbps leaving the edge. That arithmetic, not clever software, is what separates an operator that holds up during a busy evening from one that does not.

1. 1Encoding cost is per channel; delivery cost is per viewer
2. 2HEVC roughly halves the bitrate of H.264, so it halves delivery cost too

As buffering that arrives at predictable times and affects predictable groups. Oversold capacity does not degrade politely — it produces stalls concentrated at peak hours and often confined to whichever channel groups sit on the loaded edge. Two symptoms are diagnostic. If a whole category fails while others play normally, the fault is on the operator's side, because your connection has no way of failing selectively by category. If the same channel fails every evening at the same hour and plays perfectly at noon, that is load, not your line. No device setting touches either case, and it is dishonest to send readers to reboot a router over it.

1. 1Peak-hour-only failure is a capacity signature
2. 2Selective failure by category cannot be caused by your broadband

Start with scope, then run the switch-away-and-back test. Scope: one channel failing is a problem with that single source; one group failing is server-side; everything failing points at your line, your player or a stale session. Then switch to another channel and back again. If the picture returns immediately, the session stalled and bandwidth was never the limitation — a genuinely saturated connection cannot recover instantly just because you changed channel. This one test costs five seconds and settles the question most support conversations spend twenty minutes on. Only after both checks does it make sense to look at speed tests, Wi-Fi placement or player settings.

1. 1Scope first: one channel, one group, or everything
2. 2Instant recovery on switch-back rules out bandwidth

A specific one, expressed as a percentage, because the percentages convert into time you can picture. 99.9% is about 8 hours 45 minutes of downtime a year. 99.99% is about 53 minutes. 99% — which sounds respectable — is more than three and a half days. IP4KTV states 99.99%. The point is not that a number is a promise; it is that an operator willing to state a figure has committed to something checkable, while one that answers with adjectives has not. Pair the question with two others: how many simultaneous connections a plan allows, and what the refund position is if the figure does not hold.

1. 199% = 3.65 days a year. 99.9% = 8h 45m. 99.99% = ~53 minutes.
2. 2IP4KTV: 99.99% stated, 1-5 simultaneous connections, 7-day money-back on plans

Four questions, all answerable in a sentence. What uptime figure do you state? How many simultaneous connections does my plan include, and is that separate from the number of installs? Which payment methods do you accept — crypto-only operations are the signal help communities flag most often, alongside refunds offered as gift cards and outright refusal of any trial. And what happens at the end of the term? IP4KTV activates in about five minutes, allows unlimited installs against 1-5 concurrent streams, accepts ordinary payment, stores no card, and renews nothing automatically. Any operator that will not answer all four plainly has given you your answer.

1. 1Crypto-only payment, gift-card refunds and no trial at all are the three named warning signs
2. 2$5 for 24 hours is enough to test a server through one busy evening