Every new managed service feels like another layer of abstraction where the vendor's support docs become your only debugger. And they're always "optimized for your success," which means they log everything except the thing you need.
So now you're:
* Opening tickets instead of tailing logs
* Waiting for "specialist teams" instead of running a quick strace
* Paying a premium for the privilege
The cold-start latency everyone talks about isn't just in the functions. It's in the support loop. When something breaks at 2 AM, you're at the mercy of a portal and a SLA that promises a response in an hour. Good luck fixing it in five minutes like you could with a shell.
And let's not forget the migration tax. Getting your data out later? That's a separate line item they don't highlight in the initial pricing calculator.
Read the contract