I think Fulthortl might be correct. It could very well have been an open phase or an open neutral. Around me two years ago a strip of 250 watt HPS roadway lights started doing literally the exact same thing. 2 out of ever 3 lamps from a string of about 100. Upon looking at the pole feeding the lightning controller I noticed a delta-delta bank, so its safe to assume 480 volt phase-phase connected ballast. My guess is that one of the phases opened somewhere placing 2/3 of the lights in series. The striking and warming lamp causes the impedance of the fixture to change and in turn voltage leading to that flickering. Guys on Youtube are to quick to speak as usual.
HPS can't strobe or spaz blink. The ignitor simply doesn't produce enough voltage to hot strike and if it did, it would arc over elsewhere in the fixture before re-striking a hot lamp. It's something only LEDs suffer due to design flaws. Just disallow it in specs and put in remedies so the cost of corrections get entirely picked up by the LED sales vendor or the coverage they get at their cost. Alongside this, prohibit silently letting "thermal management" from compromising performance during normal usage unless it is disclosed for consideration (i.e. thermal management for parking lot may go into protective dimming if they're left on with the sun baking on the fixture)
Lighting designers or end user's buyers are not expected to intuitive sniff out specific risks that plagues certain technology but if they're given an idea of what to look for, they can just drop it into spec requirements. I place larger share of blame on the fixture integrators with the rest on the transistorized LED ballast manufacturers. It's becoming obvious to me some aren't too concerned with putting out inadequately validated beta products into major deployment.
LEDs have been extremely expensive and this is one of the reason they attract sales dealer who promise savings projection that does not happen for 3-4 times the duration of time they've been in business and by people that come from damn sales background.
Anyways being on alert of the nature of high vanishing (bankruptcy) rates of LED fixture assembly business as well as LED sales and installation and the specific failures to be on the lookout for provides clue for what to put in specs and reasonable justification for more rigorous conditions. Many LLCs have popped up to focus on those criteria. Without the proper justifications, someone will cry criteria are unfair! All the specs reqs do here is disallow certain failure mode, how it must be addressed, who will pay to have it addressed, how to keep high risk sales vendor from bidding in the first place or the type of coverage they're to provide against delayed discovery. Additionally, logically unsupportable items to avoid such as "shall be LED" which can force to select and spend more money in order to meet that specific requirement.
What this does is setup negotiated terms that could actually affect the bid price or give second thoughts to back-end issues so that LED benefits can not be sold for profit while leaving uncertainty risk with the end-user and the uncertainty is still quite high with LEDs. I heard consumer grade LED brands don't always ask for bad lamps back which is a form of acknowledgement that legitimate LED product blowouts are common enough... like LEDs.
Risk conscious companies will go up to where they get their products from and inquire them about it. Long established major businesses have to look further. his cuts them from being able to get cheap junk or very near future hopes of cost reduction engineering change because if they pitch LEDs to make money now with promises of future savings but certain items can come back and get them even well after the warranty. If spaz blink isn't specifically banned, the products that were supplied for pilot test that did not suffer form spaz blink can be made with spaz blink vulnerable ballast design in order to reduce cost. A term like "shall not fail catastrophically" will lead to disagreement over what constitutes catastrophic failure.
Decisions are often made by people who have no idea about lighting technology. LED lighting is a huge fad like many eco and green stuff and this needs to be discussed.
We're not talking about life support system or nuclear reactor design where a failure can be expected to cause catastrophic property loss or threaten life. Questions like projected maintenance cost on an experimental beta product against a proven technology is absurd. We're very aware transistors routinely fail closed while vacuum tubes seldom do or some incapable of failing that way by design. Only certain things respond well to accelerated testing. Chocolate degrade even at room temperature and the temperature affects the degradation rate but accelerating testing above certain temperature will cause it to melt which will not happen for a century at normal temperature and limits how quickly it can be accelerated.
