Newton solver stagnates for linear elastic OpInf problem

[ikalash]

In creating the clamped example for testing the new OpInf capabilities, I noticed that the norm of the residual is stagnating within the Newton solve for some test cases, e.g., the clamped problem: https://github.com/sandialabs/Norma.jl/blob/main/examples/ahead/single/clamped/dynamic-opinf-rom/clamped.yaml .

Here is how the output looks like:

Reading simulation file: clamped.yaml
Initializing run at stop 0 with time = 0.00e+00
Advancing from stop 0 with time = 0.00e+00 to stop 1 with time = 1.00e-07
|R|=0.00037316127531427496
|R|=0.0001406823671116441, solver iteration=1
|R|=0.00012675266513045592, solver iteration=2
|R|=0.00012675266513045592, solver iteration=3
|R|=0.00012675266513045592, solver iteration=4
|R|=0.00012675266513045592, solver iteration=5
|R|=0.00012675266513045592, solver iteration=6
|R|=0.00012675266513045592, solver iteration=7
|R|=0.00012675266513045592, solver iteration=8
|R|=0.00012675266513045592, solver iteration=9
|R|=0.00012675266513045592, solver iteration=10
|R|=0.00012675266513045592, solver iteration=11
|R|=0.00012675266513045592, solver iteration=12
|R|=0.00012675266513045592, solver iteration=13
|R|=0.00012675266513045592, solver iteration=14
|R|=0.00012675266513045592, solver iteration=15
|R|=0.00012675266513045592, solver iteration=16
Advancing from stop 1 with time = 1.00e-07 to stop 2 with time = 2.00e-07
|R|=0.0003157990882159685
|R|=0.0001734770428552987, solver iteration=1
|R|=0.0001734770428552987, solver iteration=2
|R|=0.0001734770428552987, solver iteration=3
|R|=0.0001734770428552987, solver iteration=4
|R|=0.0001734770428552987, solver iteration=5
|R|=0.0001734770428552987, solver iteration=6
|R|=0.0001734770428552987, solver iteration=7
|R|=0.0001734770428552987, solver iteration=8
|R|=0.0001734770428552987, solver iteration=9
|R|=0.0001734770428552987, solver iteration=10
|R|=0.0001734770428552987, solver iteration=11
|R|=0.0001734770428552987, solver iteration=12

It is very strange given that the test case is linear elastic.

Despite this apparent lack of convergence, the solution looks good for the displacement (velocity and accelerations are 0 due to an issue with their output, so can't be checked yet; see #55 ).