At the moment, GstBin creates the threadpool for state recalculations with a maximum number of 1 thread. There seems to be logic in gst_bin_recalc_state to handle the case where multiple threads enter the function (it sets a polling flag), so it seems reasonable that the threadpool should allow more threads - otherwise an application with multiple toplevel bins cannot have them all recalculating themselves in parallel.

Filing a bug because either Andy or Wim should check my logic is correct.

If it is, we should set the maximum in g_thread_pool_new to -1, and probably take the bin lock and check ->polling is FALSE before bothering to push a new recalc thread onto the pool.