I wouldn't be surprised if sed was faster than Perl with trivial regexes, because all those fancy features come with complexity and are slowing the engine down.

But one off those features is Trie optimization° of "or" alternations with |

Hence a one-pass solution with a hash lookup like demonstrated by Anomalous could be far faster in Perl than in Sed (does Sed even allow hash lookups?)

But this also depends on details only you know, like if you expect a snippet to be processed multiple times.

Like "hidings > hiding > hide" ˛ wouldn't be possible with this one-pass approach, you'd need to repeat it till it doesn't replace anymore.

updates

°) see how-does-perls-regex-implementation-makes-use-of-tries#57484188

˛) i.e. multiple reductions gerund after plural