Thanks ! I found 18 samples of delay. It seems very small, but in this case, it's enough to cause that kind of problems.

Back to the possibility of RME adding a setting in the driver of "Report Digital Latency" (basically just the ASIO latency), or "Report Analog AD/DA latency" to the host (ASIO + AD/DA conversion latencies).

At least this way with the "Report Digital Latency" setting, you will never get a "negative" value (even with a fast Lynx/Apogee converter), and loopback recording wouldn't come back early, either.  Other users apparently use "per channel" VST Delays for this task when using external Analog Inserts and such with varying AD/DA converters (RME is usually the slowest converter, so the rest come back "early" as you say), but that sounds like going back to the pre-historic days of PT LE and its cumbersome manual delay compennsation mess (no thanks!).

This seems to be coming up a lot lately   Is RME listening?

Agreed, but assuming a negative delay would be required for an analog I/O loop does seem somewhat counter-intuitive from Steiney's view IMO.

The Loopback coming back early also requires some manual user intervention, and the ping tool would not help with a re-recorded loopback track.  At least if the RME Driver would report Digital latency (as a user selectable option), Loopback would be close to accurate (1 sample off IIRC due to the routing) and external Analog Loops would be able to be measured with the Steinberg ping tool with a positive offset.

A re-write of ASIO sounds good, but not likely to happen (and then the $$$ DAW upgrade fees for all users across all platforms).  So address the problem where it is somewhat accessible - at the driver.

Do other DAW Apps have this same issue when using "quicker" converters - or do they handle negative delays more gracefully?  Just curious if addressing this at the driver with a selectable option might benefit more than just Steinberg users?

Rock on!

Those that use their ears with more complex setups involving multiple AD/DA's which must be aligned to avoid comb filtering   I believe every thread I've seen started on this topic begins with someone mentioning audible comb filtering as what brought the issue to light (not that the end-users are measuring and reporting based off the measurements alone).

I'm personally not affected by this exact issue as I use the MADI card (so no chance of my AD/DA signals coming back "early" that I'm aware of), but it would seem this might be a benefit to some more advanced users.  I know the Digiface/Multiface driver will obviously change what it reports based on which BOB is connected, so how much harder would it be to make this a user-definable setting (that defaults to Analog offsets for most user's needs) along with the "Lock Registry" and "Optimize Multiclient Operation" options on the HDSP panel's back page?

I do agree this issue does need to be addressed at a higher level (revised ASIO v3 or something similar).  But if a simple GUI add-on can solve the issue for a handful of customers (allowing positive ping offsets), then I don't see a huge drawback.

Either way, I'm still happy with all of my RME gear :-)