MS problems are continuing
On the one hand, I like that Microsoft is pushing out updates faster now (remember when security issues in Internet Explorer would sometimes takes months to be fixed?). On the other hand, it is becoming increasingly difficult to keep up with the frequent changes and what they break. 
The worst case is always one patch is made, but you need 3 others to fix the “errors” of that patch.
(This happened I remember, when had to use MS)
But the others are not better! (incl LINUX) 
The answer is open source.
MS problem is they cannot make a perfect product as then none would buy the next version to fix the planned obselence. That then means the next version has the prior planned obselence removed, the next planned obselence added, and with the change all the new bugs they not realized they put in.
We’ve got far more important things to do like get the Moon base, Mars colony, asteroid mining and explore the galaxy, and not be suffering antiquated purchasing models which distract just for a shareholder to earn $.
Too many application developers, not enough scientists.
Interesting kind of view.
We should get the problems here on our little world solved first.
Not wasting money in moon- mars and venus trips.
What about Aldebaran! 
No Cloud Service is immune to these outages… Slack had it this week, almost full day… Microsoft had it last week, Zoom and Webex have it from time to time.
there is nothing kind of stable… be it On-Prem or Cloud… rain comes down the clouds 
Cant solve anything here once the asteroid comet virus basalt gamma-ray global-warming hit.
To infinity and beyond!
Isn’t the cloud inherently less reliable?
More things to break so breaks more?
I’m aware of the shared-nothing paradigm but you’re still sharing more like sharing the same Azure data centers, same “oops I didn’t think of that” person.
No this is more reliable as ain’t broke don’t change applies, change management, outage management, go/no-go, back out. In a cloud model someone else wants a change you don’t need but you suffer it without being asked. In a cloud model you may be perfectly happy and in what on-prem could be a change-freeze as critical time of year and because someone else wants a change then tough luck.
There’s a reason you only get a cloud credit for a huge outage as they admit it is inherently unreliable.
We overlapping and effectively said the same thing. Agreed.
I do always say : Cloud is very similar to the German “geklaut”
from what I have been seeing - cloud is less reliable than on-prem, these technology companies keep pushing updates, customers have zero control on what gets pushed down the customer’s throat, lot of issues to change of functionality, and plus these outages…
Going full cloud, and then being hit with the Slack or MS kind of outage can bring the business to a stand-still.
For on-prem, there was a better control over versions, features, and failover options…
Cloud… Cloud everywhere… rain rain … oh no, its endusers’ tears actually.
Exactly! 
How many decades of work hours have been wasted by those poor users searching for errors not
caused by the customer.!
But no one listens… everyone is driven to move to the cloud.
I expect a trend for back to On-Prem, with all the outages, it is already happening, with lot of businesses thinking the “Hybrid” route, which is keep the on-prem alive
Back to the roots.
Installing a Linux-Server system is not so difficult meanwhile.
And there is much to learn and help from all over the word.
Everyone???
No. There are plenty of core business functions which never went to the cloud.
Stateless functions where multiple cached copies in shared-nothing can go cloud. Particularly if bursty load so pay just for what you need.
Statefull and business critical need a fully controlled unsourced solution with change management, change freeze.
I agree you probably need to appear to embrace cloud so the non technical people feel you’re progressive but all experienced Architects know to push unreliable-tolerant workloads to the cloud and keep reliable-sensitive in your 100% control.
Unix is still common in business critical. AIX, HP-UX, Solaris.
Yes the planet is mostly powered by Linux, it’s the most common OS in MS data centers.
COVID has actually forced more of these migrations to cloud… some core functions are still being retained on prem, but vast majority of communications / collaboration tools being moved to cloud
Is that just a Capex avoidance? If in doubt Opex?
Cloud is more expensive but a cost easier to hide - stealth IT spend.
I had a friend who went to car factory and went on a tour. He stated that while he was there they had been running an engine. The engine died while he was there and a group of the workers cheered. My friend asked the guy told that they were simulating the run time of an engine.
Right there 3 types of tests:
- unit. To test a narrow specific aspect. Many of. These are “cheap” to run and done often.
- functional. To test more complex situations, these are more expensive.
- performance. To test at full load - like your engine tested to destruction example. These are very expensive.
In an on-prem ( or co-lo, same thing just outsourced datacenter floor, power, cooling, security, everything else is in-house) you can control each change and what is tested when.
None of us wish to harm ourselves in our professional lives, so not naming anyone, I worked with a company who had two times a year when nothing could be changed, two change freeze windows. They could never hand that workload to the cloud as they could not mandate (e.g. Microsoft) to not change anything for a month. They did use cloud VMs to do their some of their testing though, the unit tests and functional tests could be run in the cloud, the performance tests on-prem, and production on-prem.
Simplistic, but those who assume something will fail, will seldom fail, those who assume nothing will fail, will fail often.
They have to know at some point “the cloud ends” and actual hardware begins right? So, like you or someone said, it is just more that can go wrong. That being said I have to give MS credit for the amount of free resources, learning modules and dev tools that they allow us access to.