[technogeek00]

The boolean tracking appears to be something like an asset state machine / tracking running in the virtual player. Manually seeking the video tag is likely an unexpected change triggering an exceptional transition in the logic.

The thing to keep in mind here is that the seeking timeline of a video tag element does not need to represent the actual media timeline when you are using MSE. It can represent any time the virtual player running in JavaScript wants to maintain and is usually more related to the buffer timeline and ensuring pts continuity across segment appends. Netflix has historically had relatively simple streams which may have allowed this to be the actual seek time, but looks like some deeper streaming changes are landing. For a more complicated example take a look at how multi-asset vod and live streams in a single video tag surface times.

I don't work for Netflix, but do work in the field and since it's relatively small know some of the folks over at Netflix. They definitely aren't out to break things purposely, it's more likely not a case they exercise and broken by a change like someone else has said. If you want to patch the functionality it may be worth exploring the dispatch of mouse clicks on the seek bar. That probably does a translate to time and then calls the virtual player seek API which would be the new hook point you want.

[spaceribs]

Yeah, I have a feeling it was due to optimizations deeper in the stack, I'm just deeply unhappy that what essentially should be an interoperable API via the <video> element was locked down in such a spectacular fashion.

The edge cases of varying video providers has been really eye opening when i've been doing this project, it hard to imagine how screenreader software (or any meta-software that makes it easier for the disabled) works at all.