Benchmarks show 54 ns (by reference comparisons) vs 1700 ns (using the internal collection comparison) on a regular Dictionary. It doesn't affect much the TryGetFullKeysByPartialKey, though, as it's more dependent on the sub-key equality comparer, which is under the callers control.
So, it's worth to try a lighter collection hashing algorithm and also check the scalability on at least some millions of records.
Benchmarks show 54 ns (by reference comparisons) vs 1700 ns (using the internal collection comparison) on a regular Dictionary. It doesn't affect much the TryGetFullKeysByPartialKey, though, as it's more dependent on the sub-key equality comparer, which is under the callers control.
So, it's worth to try a lighter collection hashing algorithm and also check the scalability on at least some millions of records.