I dunno, the TR scaling for HR makes it hard to see anything. And my over-unders don’t look like that. Carpathian Peak +2 - does it vary between 95% and 105% FTP? I did it a couple years ago using TR app (Kickr direct-drive 2017, Erg, PowerMatch) and putting it on a 95-105% color grid:
It looks similar to my outside rides at threshold.
Have you ever done a long effort at threshold? Do you know your usual threshold HR? Mine varies a little by time of day (lower in morning), cadence, and inside vs outside.
Here is a late afternoon 32-min pacing effort from earlier this year, ftp had been set at 264W and after this effort my coach bumped it up 8W.
Power view:
Slightly more variable power than inside doing Carpathian Peak +2 but basically the same.
And HR view:
My lactate threhold HR (LTHR) is pretty solid and stable since 2016, plus or minus a few beats per minute depending on a few factors (morning-vs-evening, outside temp, etc.).
A couple weeks before I did a 10+ minute pacing effort after 90 minutes of other work, it looked like this:
the instructions for that are to aim for where you perceive a 20-min TT effort to be - I did 11 min at 290W, which if it had been a 20 minute TT that would tell me (95% of 20-min works for me) to go out and shoot for a 275W long effort.
I could show longer 50-70 minute efforts, they all look similar. These benchmark efforts are done a few times a year and usually preceded a week or two by a shorter 10+ minute pacing effort to mentally reset what it feels like to do a longer ftp pacing effort.
Going back to this question, here is a similar workout from 3 days ago, its a 4x8-min at 92% with 4-min rest between:
Its 92% and as expected doesn’t hit LTHR, with recoveries taking about 45 seconds to drop back to upper zone2 HR. Normal cadence.
I reliably see this type of HR response, inside or outside, but of course its flat around here, plus access to good long uphill options, which makes it easier to do “erg-like power outside.”
What I learned in 2016 - get familiar with how your HR responds to longer threshold efforts, and use that to establish your LTHR and judge shorter efforts. This has been my human intelligence (“HI”) FTP detection and has worked well for me for years (and for 3 years cross-referenced and confirmed by Garmin Firstbeat ftp estimation which I receive at least 4 updates a month).
Hope that helps.