This time focusing on the 1.5 to 3.5 Hz band to resolve the structure around 2.6 Hz.

The noise injection ran between 1373756179 (2023/07/18 22:56 UTC) and 1373758142 (2023/07/18 23:28 UTC)

Not sure if what we see is real, but here's a fit, in s-domain, of the plant (optics plus suspension, with the controller divided out)

z = [-0.33515061+16.35437768j, -0.33515061-16.35437768j,
        -0.88008293+11.64561682j, -0.88008293-11.64561682j,
        -0.60031816 +3.15253939j, -0.60031816 -3.15253939j,
        -0.74195821 +2.5870795j , -0.74195821 -2.5870795j ,
        -0.70804008 +0.43354112j, -0.70804008 -0.43354112j]
p = [-0.57517567+17.98772839j, -0.57517567-17.98772839j,
        -0.11588748+16.31442876j, -0.11588748-16.31442876j,
        -0.55863455+15.02245062j, -0.55863455-15.02245062j,
        -0.19628544 +6.5741992j , -0.19628544 -6.5741992j ,
        -1.37239773 +2.79248047j, -1.37239773 -2.79248047j,
        -0.33923225 +2.70011481j, -0.33923225 -2.70011481j,
        -1.05551599 +0.j        , -0.12652314 +0.j        ]
k = -126469.31513966424