Yes, it's possible to run experiments modifying the pacing gain.
For BBRv1 you can modify bbr_pacing_gain, recompile, boot, and test. Or convert it into a module parameter and modify it dynamically.
Or use BBRv2, where it is already a module parameter (see
this spot in tcp_bbr2.c).
Note that you would want to make compensating changes in the first and second pacing gain of the cycle, so that the average pacing gain is still 1. (e.g. (1.25 + 0.75)/2 = 1, and also (1.3 + 0.7)/2 = 1. Also please keep in mind that larger pacing gains cause larger queues (if the bottleneck link is full) and thus larger queuing delays and potentially larger loss impacts on cross-traffic.
best regards,
neal