ValueError: b'svddc failed in l2fit when using flavor = 'seurat_v3'

• [x ] I have checked that this issue has not already been reported.
• [ x] I have confirmed this bug exists on the latest version of scanpy.
• (optional) I have confirmed this bug exists on the master branch of scanpy.

Selection of highly variable genes works fine in default settings, but I get an error when I try to use seurat_v3 flavor.

adata2.layers["counts"] = adata2.X.copy()
adata2.raw = adata2  # keep full dimension safe
sc.pp.normalize_total(adata2, target_sum=1e4)
sc.pp.log1p(adata2)
sc.pp.highly_variable_genes(
    adata2,
    flavor="seurat_v3",
    n_top_genes=3000,
    layer="counts",
    batch_key="Sample",
    subset=True
)

ValueError                                Traceback (most recent call last)
<ipython-input-18-64d280f5029c> in <module>
      3 sc.pp.normalize_total(adata2, target_sum=1e4)
      4 sc.pp.log1p(adata2)
----> 5 sc.pp.highly_variable_genes(
      6     adata2,
      7     flavor="seurat_v3",

/data04/projects04/MarianaBoroni/lbbc_members/lib/conda_envs/diogoamb/lib/python3.9/site-packages/scanpy/preprocessing/_highly_variable_genes.py in highly_variable_genes(adata, layer, n_top_genes, min_disp, max_disp, min_mean, max_mean, span, n_bins, flavor, subset, inplace, batch_key, check_values)
    417 
    418     if flavor == 'seurat_v3':
--> 419         return _highly_variable_genes_seurat_v3(
    420             adata,
    421             layer=layer,

/data04/projects04/MarianaBoroni/lbbc_members/lib/conda_envs/diogoamb/lib/python3.9/site-packages/scanpy/preprocessing/_highly_variable_genes.py in _highly_variable_genes_seurat_v3(adata, layer, n_top_genes, batch_key, check_values, span, subset, inplace)
     83         x = np.log10(mean[not_const])
     84         model = loess(x, y, span=span, degree=2)
---> 85         model.fit()
     86         estimat_var[not_const] = model.outputs.fitted_values
     87         reg_std = np.sqrt(10 ** estimat_var)

_loess.pyx in _loess.loess.fit()

ValueError: b'svddc failed in l2fit.'

Versions

0.10.00

Start by reproducing the shown sc.pp.highly_variable_genes call with flavor="seurat_v3", layer="counts", and batch_key="Sample". Read scanpy/preprocessing/_highly_variable_genes.py around _highly_variable_genes_seurat_v3 and the _loess.pyx l2fit failure; done means the reported input no longer raises ValueError.