Not familiar with gnm, I suspect the problem is that stratum_YMD only has as many levels as values. When you set it as eliminate, you are including it it the model as an independent variable, so the fit fails (because each point has a different value).
You can see that by looking directly at the fit:
length(test$stratum_YMD) == length(levels(test$stratum_YMD))
#> [1] TRUE
one <- gnm::gnm(Lost ~ HIgroup1 + PRgroup2 , data=test, family=poisson,
eliminate = stratum_YMD, exponentiate = TRUE)
one
#>
#> Call:
#> gnm::gnm(formula = Lost ~ HIgroup1 + PRgroup2, eliminate = stratum_YMD,
#> family = poisson, data = test, exponentiate = TRUE)
#>
#> Coefficients of interest:
#> HIgroup12 HIgroup13 HIgroup14 PRgroup22 PRgroup23
#> NA NA NA NA NA
#>
#> Deviance: 4.849983e-08
#> Pearson chi-squared: 2.424991e-08
#> Residual df: 0
one_no_stratum <- gnm::gnm(Lost ~ HIgroup1 + PRgroup2 , data=test, family=poisson,
exponentiate = TRUE)
one_no_stratum
#>
#> Call:
#> gnm::gnm(formula = Lost ~ HIgroup1 + PRgroup2, family = poisson,
#> data = test, exponentiate = TRUE)
#>
#> Coefficients:
#> (Intercept) HIgroup12 HIgroup13 HIgroup14 PRgroup22 PRgroup23
#> 2.43792 0.12085 0.33983 0.26538 0.04811 0.14696
#>
#> Deviance: 526.6366
#> Pearson chi-squared: 500.3253
#> Residual df: 123
Created on 2020-12-07 by the reprex package (v0.3.0)
So, are you sure that stratum_YMD is an appropriate variable to include?