I've tried different ffv1 options on a 2,4GiB (time=02:36.24, 2 and a half minutes) file and wondered whether it might be wise to enable the following options by default, as they are all supported by ffv1 version 1 (stable since March 2010):
1) use range coder (option coder=1, yields ~152 MiB)
2) set context to large (option context=1, yields ~65MiB)
On my machine (Core i7-2600k), throughput dropped by about 5-7 fps when enabling both options, but since ffv1 is lossless and for archival purposes maybe enabling them by default is worth it?
Next there is also ffv1 version 3 (option level=3), released in August 2013, which supports multi-pass encoding. However, with this file multi-pass encoding did not yield very much (that is, only 12 MiB) in terms of compression, especially when put in relation to the time needed for the extra pass. I still have to evaluate whether version 3 is faster than version 1 in non-multipass encoding.
More info about the various options is available here (for example): http://www.ffv1.info
g=1, coder=1 and context=1 are added to the lossless/FLV1 preset in git commit d1a9780. We do not set codec-specific defaults in the code.