Some developers have referred to this as an ugly hack, is it accurate?

It was originally thought that segregated witness could only be added to Bitcoin using a hard fork. For that reason, it was considered unlikely to be possible to add to Bitcoin in the near term because hard forks are much harder to gain ecosystem-wide consensus for.

However later on, it was discovered that segregated witness could be added in backwards compatible manner as a soft fork. This method may have been described by segregated witness’s designers as a “hack” but this is more a colloquial term among programmers that refers to achieving something thought near impossible in a pleasing way the aesthetic appearance: the actual difference is simply between the location of the witness commitment (in the block header for the hard fork version or in the coinbase transaction for the soft fork version).

In fact, the net effect of segregated witness actually cleans up a lot of technical debt rather than adding to it. If there ever a hard fork in the future, the location of the witness commitment could be relocated as was mentioned in the original roadmap proposal. It is worth pointing out that the hard fork implementation involves its own tradeoff by adding 32 bytes every header required for SPV wallets to operate.