So I refrained from responding to some messages that I read in here regarding which I had some thoughts or own opinion, and I wasn't involved in these votes (and probably will not vote in this case), but there's one thing I found worth mentioning at this point regarding the types and availability of (submitted) movies, maybe as a tip for future TAS submissions (or creations of them).
Namely there's something that I'm slightly worried about regarding the approach that HappyLee took in this TAS instance, and in order to explain what I mean, I'll go through a thought experiment:
Imagine you'd have some game given together with a wish/task to find/make the fastest possible TAS (let's say for a given fixed branch), and after spending a lot of time on the game, you'd end up with some (pre-)finished movie file that (expectedly) demonstrates some approximation of the best possible time (note here that the best possible time is also an approximation of itself, so this case shall not be excluded here). And let's say that at this initial stage, the focus so far entirely lied on the (for TASVideos) primary goal of reducing the total time of the TAS, and further variations, changes, maneuvers, playaround, meant for a secondary goal such as entertainment remains to be worked out and implemented into the existing (pre-)finished movie.
Now, if you don't know yet if the TAS could still be improved with respect to the primary goal (total time) or not and you go over to the/a 2nd stage and apply changes to the inputs of the TAS in such a way that they (possibly) just barely anymore fulfill the conditions needed to still meet the same finishing time of the movie that you initially managed to cut your TAS down to, then (now from a community perspective, a perspective that wants the strive for optimality) anyone else in the future that would want to continue the work on this game for a potentially possible improvement might now have to put in more work, possibly because said future person might be left with the initial ''burden'' to take out all the (with respect to the primary goal) unnecessary slowdowns/detours/variations (or in whatever form or shape it happens to exist) from the current existing TAS in order to get back to the ''clean'' situation (i.e. the form of the movie file as it was at the end of the first stage of the TASing process, before the entertainment variations were implemented) that could (for such future TASer) possibly be expected to be in a closer situation (than it might be for the entertainment-filled TAS case) to a situation that would cut off more time (provided it exists). By ''closer'' I mean a difference regarding the amount of inputs needed to change to get to a faster TAS.
As an example that let's say involves a frame-rule such that time improvements only matter in multiples of 20 frames within which one gets to some point in the TAS, so that the same time would be obtained if one gets there within 5 or 17 frames, and would only be one further 20-frames long step slower if one would instead get there at the 21th frame or the 40th frame:
With a given entertainment TAS that barely ''makes the cut'' to clear the game in some 20 frames interval I'd expect it to be more work from there to cut off further frames*, since at first, the ''improvements'' made by a future TASer will then not yet allow to make any difference time-wise and might have been known already by someone else, and said future TASer might at this point not yet know if he/she actually found a new improvement or just erased an old purposefully implemented slowdown out; compared to if one started from a TAS that is ''clean'' and expected to be closer to the next frame-interval to be one step faster.
* With these frames of which I expect them to be easier cut off, I refer to frames that can be cut by grinding out close variations of given inputs for a segment (i.e. frames for which I expect that a brute-force would have it easier to cut them out), rather than frames that are cut due to a new idea/glitch being involved (and for this latter type of frames that could still remain as frames waiting to be cut, it wouldn't be as clear to determine which TAS version, a ''clean'' one or an entertaining one, would be seen as being ''closer'' to finding these or inspiring to find these).
So for the greater general benefit, I'd expect it to be useful in general (if possible, since the circumstances might varii from game to game and TAS to TAS) if one would/could just provide multiple files (maybe alongside a submission, or in a corresponding game discussion thread), namely a ''clean'' version that is meant to be a new foundation for further experimentation and input variations for new cuts, and then (if it exists/was done this way) your personal (entertaining) variation of it (and maybe even a 3rd version that contains unnecessary but potentially helpful e.g. 2nd controller inputs as instrinsic markers in an input editor window for future TASers for orientation and to immediately see when some events start or end or other ''sidenotes'' that one can insert into movie files via inputs).