Any predictions on the impact of the 'Y38' bug (32bit EPOCH time issue)?

Y38 is really a non-issue. It only involves using more than 32 bits in the epoch-counter. (Strictly speaking, merely moving from 32 bits to 33 bits would take us to 2106 or thereabouts.)

The easiest way to fix this for millions of years would be to switch from using a 32-bit counter to using a 64-bit counter.

You can be sure than entities requiring to use dates later than 2038 today (banks, insurance companies, etc, etc.) are doing this already, others who don't need to do that yet probably won't bother doing that until they actually need to.

Very little programming work is required. In many cases, it requires little more than recompiling after updating a few definitions.