GAMGAgglomeration unexpected caching behavior for multiple solvers

When running a simulation where multiple equations are solved/preconditioned using a GAMGSolver, it seems like the caching of the GAMGAgglomeration is shared. This means that the GAMGAgglomeration is only done once (by the first equation's solver), and this cached Agglomeration is then also used by the second equation's solver. However, I am trying to use different agglomeration settings for both solvers, but now the settings for the second solver are ignored since the cached version is used... Of course I could turn caching off but this will worsen the performance of the simulation.

Is this behavior by design, or could it be a bug? Maybe I am missing something or misunderstanding parts of the code. I was hoping someone can comment on this behavior and discuss whether caching should be supported per equation (or even per setting, such that caches can be shared if the same mesh + settings are used).