Leaving status as Noted – see https://blogs.technet.microsoft.com/configmgreng/2016/03/11/configmrguv/
We don’t do anything explicit to block GPO processing, in earlier Windows versions that wasn’t the case but GPO apply was problematic in some instances
Our next step for this item is clear explanation of how GPO relates to OS Deployment Task Sequences and standalone Task Sequences.
Basically a lot of configuration is done via GPO that can be quite critical during OSD.
Main example: Bitlocker.
Quite often you have a Bitlocker GPO defined that says for example "Encryption is AES-256, store TPM password in AD".
However, during OSD, as this GPO doesn't apply, the TS will just use the default settings of Windows, which in this case would be enabling Bitlocker with AES-128, which isn't what you want.
The current workaround is manually using "run commandline" steps in your TS to set the exact registry keys you set with that GPO. That's redundant work and can be error prone.
There are many other examples like this too.
Beyond that, by default after OSD finishes GPO hasn't applied. You need to use some _SMSTSPostAction to do a gpupdate /force and reboot to be even closely sure that GPO will be applied. This is important in high security environments where GPOs are used to enforce a Secure Configuration Baseline such as the CIS benchmark.
yes, lack of control of how dependencies are installed (or just control of general order of installation) is why we are not using App model right now and are sticking to TS and Packages.
Or just add a "process Group Policy" step to Task Sequence.
Updating status to planned.
We shipped our 1905 Technical Preview today.
This preview contains the first release of the task sequence debugger. We’ll continue working on it to address more of the asks in this item.
Quite often problems a TS encounters are "fixable", but you can only completely restart a TS from the beginning, not say "restart from this step" or "restart from the failed step".
It would be nice to have such an option, as it would for example allow you to log on, investigate, fix the possible issue and basically tell the TS to continue where it left off.
Similarly, you could force the TS to maybe halt execution till you resolve the issue, and only continue after a time-out or after you give a specific instruction.
By extension, just allow TS to be deployed to user collections. I love "user-centric" stuff, but I hate that I can only send applications user-centric. TS are insanely powerful tools, but I'm forced to do it machine-based.