I checked the scene, here are a few tweaks to help in your scene:
1) Light strength
Is very unbalanced in your scene! The "Directional Light" material on the spot lighting the rights side of your room is so bright (but very far away) that it eats up close to all computation time of your GPUs!
Set the "Em. Sampling Mult" of it down to 0,00001
, and you'll end up with better weighting.
Indigo has a simple but usually very functional algo to weight computation time over all the emitters in the scene:
The brighter the lightsource, the more important it is and the more it does contribute to the scene.
This does work well in most cases, but can be way off in some like yours too. You have a very bright emitter very far away. This emitter is by far so much stronger then all the other emitters in the scene, that Indigo dedicates close to all computation time to it, causing the other emitters to be so slow that they are super noisy...
The best way to avoid such traps is to work with LightLayers during scene creation. By that you can check every Emitter for itself and adjust for example the "Em. Sampling Mult" to boost its priority if its causing more noise then the other lightsources. For the final rendering you can put all emitters back to the same LightLayer....
2) Prevent Fireflies
Try to combine Clamp Contribution of 5 with some Super Sampling that can help eliminating FFs very well to as long as you don't work with untonemapped EXR output in your pipeline. The bigger your emitter meshes are the less noise they will cause since they'll get easier to find by the Path Tracing algo.
Your Floor has a Fresnel scale of ~0,062, by that you'll tune it way done killing about all reflectivity it had from the IOR of 2,0
Here a result after raising SuperSampling to 3, Clamping to 5 and adjusting the one emitter to lower multi, 2,24min on my 1080Ti:
For denoising animations I highly suggest a temporal denoiser that compares multiple frames to each other and by that defines what is noise and what is scene detail, I use Neat Video for that.