Sequential voicing is hard to predict. While it can occur in a compound (yes, it always occur for compounding), it is not mandatory. For other compounds it is obligatory. I am not sure if the preference or obligation is influenced by dialectal* differences as word accents are but this thread so far points at that possibility. *I used the word dialectal as a nifty expression for any differences that a speech community of a language such as Japanese entails. This idea involves but is not limited to regional differences.
ころ vs ごろ
If the word makes a compound with a regular noun, sequential voicing is very likely to occur.
Examples are; いまごろ, あきごろ, けさごろ and so on.
このごろ does not belong to this category of voicing since この is properly an adjectival (that is, not a noun).
If an adjectival component (another small letter caveat is that adjectival includes all components that modifies ころ as a noun but not limited to the part of speech called "adjectives" in regular pedagogical grammar) is used to make a phrase with koro, sequential voicing is avoided (with perhaps a few exceptions such as このごろ).
Examples are; まだくらいころ, あのころ, このころ, ちいさいころ and so on.
このころ is more or less like "this period of time" and can refer to events in the past and the future. Sequential voicing narrows このごろ down to a short strip of time that includes the time of utterance ("lately" or "these days"). It is an idiomatic exception to the adjectival class.
If koro makes a compound with a numerical expression, sequential voicing is likely to occur.
Examples are; さんじごろ, じゅうがつごろ, じゅうにじにふんごろ and so on.
とおり vs どおり
とおり in sense of "street" always undergoes sequential voicing.
Examples are; おおどおり, こまちどおり, そとぼりどおり and many many others.
I am not sure if numerical expressions can be used with this sense of とおり.
If the word is used in sense of "manner" or "method," again, sequential voicing is sure to occur.
Examples are; マニュアルどおり, へいねんどおり, よそくどおり and so on.
とおり can also be a counter of "manners" or "methods." In the usage, it hardly (read; probably never) undergoes sequential voicing.
Examples are; ひととおり, ふたとおり, さんとおり and so on.
くらい vs ぐらい
Kaori has already mentioned that sequential voicing is optional but frequent. I also think くらい and ぐらい are free variations. That is, they can replace each other without causing the sentence to be ungrammatical. It will be great, however, if kind people in this forum could help find expressions that avoid / prefer sequential voicing. They would either be idiomatic exceptions such as このごろ or cast more light on the semantics of くらい.