Lloyd is very smart and you'll often see deeper insights from him than most other reviewers, and certainly more than a superficial vlogger-style review of the iMac Pro.
He makes some good points about Apple's numerous quality control issues in recent years. E.g, the dumbed-down Disk Utility which actually had a broken UI in El Capitan -- the cursor turned to a column separator but the columns wouldn't move.
On the iMac Pro I had a similar "infinite boot loop" problem as what Lloyd mentions. This isn't unique to the iMac Pro and has been periodically reported by other users on various Macs:
https://techtalktone.wordpress.com/2013/09/16/stuck-in-boot-loop-on-the-mac/
I booted to safe mode, did a few things recommended in the above article and it finally started working. Obviously that should never happen but it's not unique to the iMac Pro.
Re performance of the iMac Pro vs the top-spec 2017 iMac, I have both of those on my desk, and the iMP is 10-core Vega 64 with 64GB RAM.
One some workflows the iMP is a lot faster than the top 2017 iMac. On others it's not. Lloyd is correct that it's not always faster, but this isn't necessarily Apple's fault and has little to do with the iMP design. A 10-core Xeon PC might not be any faster either on those workflows.
The sad truth is performance evaluation is now much more complex and much more limited by application software design. There are a myriad of software "load paths" through the OS and hardware space. One path can be fast and the other slow -- on exactly the same hardware. It depends on how the app software leverages the underlying features. If it uses efficient multithreading, if it properly uses hardware accelerators like Quick Sync, nVidia's NVENC, AMD's VCE and AVX vector instructions, it will be a lot faster on new hardware with those features. If the app software does not properly leverage those, the expensive new hardware won't help much.
There can even be great variation within the same app. E.g, on a 10-core Vega 64 iMP, FCPX exports to 4k 10-bit HEVC 32x slower than 8-bit HEVC. That is likely because the current software and OS don't support hardware accelerated 10-bit encoding on the Vega 64 VCE logic. You could maybe blame Apple -- from a software standpoint -- but that wasn't Lloyd's point. His point was the iMac Pro hardware. Yet from a hardware standpoint, this isn't Apple's fault. They are constrained because Intel doesn't make a Xeon with Quick Sync.
In most Lightroom 7.2 tests I've done between the 10-core iMP and top 2007 iMac, LR isn't that much faster. It's about 30% faster on import and 1:1 preview generation but no faster on many other tasks. Like the regular iMac, the iMP often will speed up in LR if you disable LR GPU acceleration. This is a profound coding problem on Adobe's part.
In Photoshop one of the most powerful (and CPU-intensive) filters is "shake reduction" sharpening. However I don't see any performance improvement on the 10-core iMP vs the top iMac. If Adobe is going to keep writing software like that, you can't fix that with hardware.
The iMP is a lot quieter than the i7 iMac when under sustained high-CPU load, so that's pretty nice. You don't need a 10-core machine to obtain that -- the base model iMP ($3999 at Micro Center) gives that, which is only a few hundred $ more than the top-spec iMac.
