One other thought that cropped up this morning was this though - as the Tesla cars themselves are so software-centric, I hope they are far more rigorous and disciplined on quality control for their vehicle software than they are for their website
My TD;DR; would be:
Infotainment QA is lousy. Bugs are, generally, fixed within a few releases (couple of months) of a major release. But old bugs, fixed, reappear in future releases ... generally the direction is forward though ... and looking back on 3 years of ownership I have all sorts of new features that weren't even alluded to at purchase (including a complete restyling which I hated - change of buttons to "flat" style, which my aged eyes struggle with ("Not the car I bought"), and the latest which changed the way "windows" work which I think is clever because it makes it much easier to rearrange screen real estate for new features) .
I mean, "fart mode". What's not to like? I'll make a prediction that won't appear on iPace
For example: Browser became all but unusable with launch of V9. Turned out that (I presume because a bit of the map now pokes out of the top of a "full screen" APP) this was related to Live Traffic, and if you turned that off Browser was OK. They've fixed that now ... but Automated Regression Testing would have picked that up and should be standard in a software organisation as big as Tesla - heck, my company is tiny and even we do that.
But it was quite some time until someone spotted the correlation ... and if you didn't read this forum (and trawl all those threads) you would have no clue as to why.
Safety-critical side QA seems to have been good. There have been fatal accidents on AP but (in all the ones I have read) it seems the Driver had ample time to do something, so was either medically incapacitated or not paying attention. They are YouTubes of people driving the same fatal road stretch, using same version of AP, car drove the same, incorrect, line ... but driver took over, and had ample time to do so.
My plea to newbies: never, ever, be tempted to "just read a text", or similar, on AP. The problem with AP is that it is very reliable. The occurrence of a life threatening event may average well less than "during one ownership", but I don't want to be that statistic ... you neither ...
In terms of Safety-critical "bugs" the only one I can think of is phantom braking (car thinks a gap is too narrow / some hazard on the road etc.) That varies from version to version and sometimes gets worse in the short term. New AP features are often released long before they are safe to use - but that's fine, try them, and then turn them off until improved. The current Navigate-on-Autopilot, automatic lane change, is in that bracket.
The problem is Humans are not expecting any of the weird things that AP / AI does. If you took a 17 year old out for their first drive you would have a pretty good idea of all the daft things they might do, and be prepared for them.
Not so with AI ... my car will think a juggernaut is a thread one time out of hundreds, and to my eye the gap looks at least as big as the previous hundred it drove past.
I drive with foot hovering over accelerator to counter phantom braking - so the guy behind doesn't run into me, because if I'm not seeing anything ahead that is a real threat he sure as heck won't be! But that leads to a further risk that AP has spotted a real danger, the most likely being car two-in-front is making emergency stop and car in front isn't showing brake lights (yet). Hitting the accelerator at that point, to override AP braking, would not be great! Good news is the moment you come off the accelerator AP will be on the brakes, before you get to them ...
The one I have seen is a Twitter feed
Model 3 VINs on Twitter