According to http://prize.hutter1.net/#history the current winner (which has stood since 2009) took 9h on a 2GHz P4 to decompress its input to 100MB, or roughly 3082 bytes/s.

30 minutes to decompress 5324821B is only slightly less, at 2958 bytes/s. This "on a modern computer", certainly not one from 15 years ago, even after factoring in JS overhead, sounds doable, if not easy. Have CPUs gotten more than an order of magnitude faster since then? Probably (but perhaps not for the type of code that would be encountered in a decompressor --- the biggest boost has been for vector operations, which are neither particularly suitable for high compression algorithms nor easily usable from JS.) Is JS overhead less than 10x native code? Probably for some cases too.

In fact, I just tried paq8hp12 on the file and it compressed it to 1059027 bytes in 208 seconds on a moderately old system (i7 870). Decompression took the same time.