Just a few initial random thoughts:

Were you able to suss out where POWO is getting their treatment? At the moment I'm not thinking of any North American references that are splitting up L. lepidus s.l. I'll have to check the draft FNA treatment when I have access to it again in a week, but I don't think it splits up L. lepidus either. Not that I don't agree that splitting may be warranted, just interested in assessing the merit of whatever source POWO is using.

POWO recognizes two subspecies of L. culbertsonii, so I think L. lepidus var. culbertsonii should be mapping to L. culbertsonii subsp. culbertsonii.

I agree that atlasing may not gain us much here, but with 521 lepidus s.l. observations about to get bumped up to genus level, I wonder if it would still be worth a try for whatever can be salvaged?

Or alternately, just move them all to lepidus s.s., and let subsequent IDs take care of them?

With that in mind, and continuing from the forum, I'm noting that there are references to a L. lepidus-group as a taxonomic concept. If this were also added as well, this could be used as a catch-all for the 521 (wherein the split would only bump IDs so far as this complex). This would address both issues of easier diagnosis of the s.l. taxon (equivalent with the species group concept) while also allowing alignment to POWO on species definitions. Complexes should generally be a rare addition, but Aaron's points above may be sufficient rationale.

I disagree with this splitting as well.
The genetic work I've performed shows little to no distinction among varieties (though I didn't have sampling for vars. ramosus, culbersonii, or cusickii), and in the field the varieties intergrade.
I also recommend following the treatment in the Jepson Manual, with all ten taxa (aridus, ashlandensis, confertus, culbertsonii, cusickii, lepidus, lobbii, ramosus, sellulus, utahensis) regarded as varieties (note that vars. lepidus, ashlandensis, cusickii, and aridus occur outside of California and aren't listed in Jepson).
Note that this differs from the current iNaturalist circumscription by including aridus, cusickii, sellulus, and ashlandensis as varieties under L. lepidus.
https://pdxscholar.library.pdx.edu/open_access_etds/919/

Just anecdotally the caespitosus-utahensis thing (https://www.inaturalist.org/taxa/241534-Lupinus-caespitosus-utahensis) has always seemed pretty distinct in the field to me. But I may not be aware of some parts of its range where it might intergrade with other variants of L. lepidus s.l. And I also don't know what distinguishes var. caespitosus from var. utahensis.

Thanks @twr61 for a very thorough summary.

Long story short, for now, I would put them in synonymy with their parent taxon (e.g., var. aridus or var. sellulus, respectively, and keep var. ashlandensis).

The "problem" is that these names generally do recognize actual, stable, morphological diversity in the field, but the morphologies intergrade to such a degree that it's difficult to draw clear boundaries, and it's unclear if there is genetic or reproductive isolation among them. There are literally dozens of named entities in this group, and most populations could be called distinct in some way.

For loloensis and lenorensis I haven't had the chance to see them, or include them in my study, so I don't have personal knowledge. From Cox's description, however, they look like they are solidly within aridus. The updated Flora of the Pacific Northwest covers this area of Idaho, and puts these in synonymy with aridus.

Ashlandensis does have a distinct morphology. Cox seems to have put it as a variety under aridus based on allozyme data; my genetic data show it to be similar to nearby populations of sellulus and lobbii.

I sampled plants from roughly the type locality of medius (near LaPine, Oregon) and called them sellulus. This area seems to be a transition zone between aridus to the north and sellulus farther south.

I haven't had personal experience with artulus or ursinus. Cox differentiates them from typical sellulus based on flower size and shagginess, respectively, but Jepson doesn't recognize them in California.

This group covers a large area spanning across the boundaries of the regional floras (PNW, Jepson, Intermountain), but they all recognize these as varieties of L lepidus. Perhaps the upcoming treatment from Flora of North America will provide a better synthesis.

Copying the text from the draft taxon split that I'm about to delete:

To conform to the Plants of the World Online taxonomy, the varieties of Lupinus lepidus are now either moved to varieties of existing species (sellulus, caespitosum) or elevated to new or former species (culbertsonii, cusickii). Lupinus lepidus sensu stricto now only applies to the former L. lepidus var. lepidus found in the Pacific Northwest. @jdjohnson has attempted to classify observations identified as Lupinus lepidus into varieties where possible, though many ambiguous observations remain and will be reassigned to the level of Lupinus.

Key to Lupinus lepidus sensu lato varieties

by Janel Johnson (@jdjohnson); last updated 6 Aug 2019. Key cobbled together from Intermountain Flora, 1989; Flora of the Pacific Northwest, 1981; Jepson, 2012; Illustrated Flora of the Pacific States, 1950.

A1 Racemes sessile or very short-pedunculate, shorter than the leaves and largely concealed by them

B1 Wing petals slender, 7-8 mm, about 1/3 as broad; banner slender, width < 3/5 length; c Ore to w Mont and s to Colo and Utah, also in White Mtns, Cal…
… var. utahensis >> L. caespitosus var. utahensis

B2 Wing petals broader, gen > 8 mm and nearly 1/2 as broad; banner broader, width 3/5 of length; Blue Mts, Ore, and Okanogan Co, Wash… var. cusickii >> L. cusickii var. cusickii

A2 Raceme pedunculate, at least partially surpassing leaves

C1 Plants prostrate and matted

D1 racemes mostly < 5 cm at anthesis; upper montane to subalpine; BC to s Cal, e to w Ida and Nev var. lobbii >> L. sellulus var. lobbii

D2 racemes elongate mostly > 5 cm at anthesis; dry, open pine forest to moist meadows, s Ore and n Cal, e to s Ida and ne Nev … var. sellulus >> L. sellulus var. sellulus

C2 Plants not matted

E1 Plants with simple or branched stems generally more than 10 cm long and bearing 4 or more leaves

F1 Racemes usually solitary and terminal, long and dense, 9-15 whorls; vernally moist meadows in pine and sagebrush; ne Cal and nw Nev …
… var. confertus >> L. confertus

F2 Racemes usually several, short and open, 3-7 whorls; rocky ridges and hillsides; ec Cal

G1 Leaflets 10-30 mm, flowers 9-11 mm, Kaweah River, Tulare & Fresno Cos. var. culbertsonii >> L. culbertsonii

G2 Leaflets 5-15 mm, flowers 7-9 mm; c Sierras, White and Sweetwater Mountains, Cal… var. ramosus >> L. confertus

E1 Plants with short stems, less than 10 cm and bearing fewer than 4 leaves

H1 Racemes gen partially concealed by longer leaves; flowers mostly ~ 9-11 mm; sandy or gravely hillsides in sagebrush and juniper; sc Wash to Cal and Nev, e to sc Ida var. aridus >> L. aridus var. aridus

H2 Racemes gen exserted well beyond longest leaves; flowers 11-13 mm; dry prairies and slopes; s BC to nw Ore, w Cas … var. lepidus >> L. lepidus sensu stricto

And just when I thought this was done.

This section (http://www.oregonflora.org/family_treatments/Fabaceae.pdf page 32) from the Oregon Flora Project lists several varieties of Lupinus lyallii as synonyms for Lupinus lepidus lobbii:
Lupinus lyallii A. Gray var. fruticulosus C.P. Sm.
IL - 2:495, accepted, in Oregon (explicit); PEK - 453, accepted, in Oregon (explicit).
Lupinus lyallii A. Gray var. lobbii (A. Gray ex S. Watson) C.P. Sm.
IL - 2:495, accepted, in Oregon (explicit); PEK - 453, accepted, in Oregon (explicit).
Lupinus lyallii A. Gray var. lyallii
IL - 2:495, (subtaxon implied), accepted, in Oregon (implied); PEK - 453, (subtaxon implied), accepted, in Oregon
(explicit).
Lupinus minutifolius Eastw.
IL - 2:631, accepted, in Oregon (explicit).

Jepson eFlora lists L. lyallii var danaus and var. lyallii as synonymous with L. lepidus lobbii.
http://ucjeps.berkeley.edu/eflora/eflora_display.php?tid=61363

Intermountain Flora lists L. lyallii ssp. lyallii, L. minutifolius, and L. washoensis and synonymous with L. lepidus var. lobbii and L. lyallii ssp. subpandens with L. lepidus var. selulus (Vol 3B pages 256-259)

There are currently 10 var/ssp of lyallii but I don't have references for the four below that show the native range in parentheses.

Lupinus lyallii https://www.inaturalist.org/taxa/164908-Lupinus-lyallii
Lupinus lyallii ssp. alcis-temporis (Idaho)
Lupinus lyallii var. danaus
Lupinus lyallii var. fruticulosus
Lupinus lyallii ssp. lyallii
Lupinus lyallii var. macroflorus (Washington and Oregon)
Lupinus lyallii ssp. minutifolius
Lupinus lyallii var. roguensis (Oregon)
Lupinus lyallii ssp. subpandens
Lupinus lyallii var. villosus (California and Oregon)
Lupinus lyallii ssp. washoensis

Insights and opinions are greatly appreciated.

Consortium of Pacific Northwest Herbaria to the rescue. Their checklist covers everything I haven't been able to find so far and agrees with Jepson, Intermountain Flora, Oregon Flora Project, and Flora of the Pacific Northwest 2 on 98.7% (only one disagreement for L. lyallii ssp. subpandens.)
http://www.pnwherbaria.org/data/checklist.php
I'm going to get all of these swaps and taxon framework references set up and I'll let them sit for a while to see if we get any more comments.